JP2003050987A - Optimization adjustment method and optimization adjustment device - Google Patents

Abstract

(57) [Summary] [PROBLEMS] To realize a method for efficiently adjusting a device considered to be optimal for a user. SOLUTION: A first step of setting an initial set of solution vectors to be updated, and setting an initial solution vector group around each vector in the solution vector set, and adapting the solution vectors belonging to the initial solution vector group to the problem A second step of repeating a solution vector rearrangement operation a predetermined number of times based on the degree, and a solution vector in a solution vector set obtained by integrating a plurality of solution vector groups obtained in the second step into one. A third step of performing a predetermined rearrangement operation based on a genetic rearrangement procedure, and a fourth step of outputting an optimal solution vector. In the second step, until a predetermined termination condition is satisfied, After the local update is repeatedly performed, the global solution vector is updated in the third step to quickly estimate the optimal solution vector.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optimization adjusting method and an optimization adjusting apparatus using a conventional genetic algorithm used for a traveling salesman problem, an optimal solution estimation problem including a circuit design problem, etc. The present invention relates to an optimization adjusting method and an optimization adjusting apparatus capable of efficiently searching for an optimum solution by adding a device for compensating for the lack of local solution searching ability which has been a drawback.

Furthermore, the present invention provides an evaluation standard including adjustment of characteristics of hearing aids, design of products and interiors according to concepts that cannot be explicitly shown, and adjustment of acoustic and image characteristics according to individual tastes. To the problem that the adjustment result cannot be evaluated quantitatively because
The present invention relates to an optimization adjusting method and an optimization adjusting apparatus thereof, which efficiently uses an interactive genetic algorithm capable of obtaining an optimal solution vector for the user based on the subjective evaluation of the user.

[0003]

2. Description of the Related Art A genetic algorithm is an algorithm for searching for an optimal solution, which has the advantages of quickly converging to the neighborhood of an optimal solution and being unlikely to fall into a local solution, and has received particular attention in the field of combinatorial optimization. There is. For the conventional technique of the optimization adjustment method and the optimization adjustment device by the genetic algorithm, for example, "Genetic Algorithms in Search Optimization and Machine Learning"("Genetic Algorithms in Search, O
ptimization and Machine Learning "(David E. Goldber
g, Addison Wesley)), JP-A-2-236660, and the like. The outline will be described below.

FIG. 68 is a flow chart showing the principle of optimal solution estimation by a conventional genetic algorithm. << Step 1 >> Consider a set P consisting of n elements based on an m-dimensional vector p _k (p _k1 , p _k2 , ..., p _km ). Each element of the m-dimensional vector p _{k that} is each element of the set P represents a specific solution of the corresponding parameter in the optimal solution search problem given as an object, and the vector p _k is a specific solution vector. Equivalent to. Each element p _ki of the vector _{p k (i = 1, ...} , m) is called a gene from the context of the organism, the vector p _k is sometimes called a chromosome. When using a genetic algorithm, first an appropriate initial set P of solution vectors is created. << Step 2 >> The goodness of the solution of each element (chromosome) of the set P is evaluated according to a preset evaluation scale, and the result is expressed as an evaluation value. Here, the preset evaluation scale is called an evaluation function (fitness function). << Step 3 >> The suitability of the solution is determined based on the magnitude of the evaluation value obtained in Step 2. At this time, due to the problem,
There are cases where a larger evaluation value is a solution having a good fitness and a smaller evaluation value is a solution having a good fitness. The former is called a maximization problem and the latter is called a minimization problem. Remove the low-fitness solution vector from the current solution set and selectively allow the high-fitness solution vector to survive. Such an operation is called selection. Various methods are known as selective selection methods, and the details thereof will be referred to the aforementioned references. << Step 4 >> The solution vector set selected in Step 3 is subjected to a genetic recombination operation such as crossover or mutation to create a new solution vector set. The number of solution vector individuals included in the solution vector set is
The value is fixed here, but may be increased or decreased. 71A and 71B are explanatory diagrams of the concept of the recombination operation in the genetic algorithm. FIG. 71A shows an example of the crossover process, and FIG. 71B shows an example of the mutation process.

Crossover is an operation of creating a new solution vector by replacing a part of a solution vector represented by a finite symbol with a part of another solution vector as shown in FIG. Further, the mutation is an operation for changing a part of the components of the solution vector selected from the solution vector set to another symbol with a certain low probability as shown in FIG. The crossover process corresponds to a search at a position far away from the current solution vector, and the mutation corresponds to a search near the current solution vector. Through these two processes, the genetic algorithm creates a new set of solution vectors. Although various methods have been proposed for the processing of crossover, mutation, etc., the details thereof will be referred to the above-mentioned reference.

In the genetic algorithm, the above processing is repeated, and as a result, each solution vector of the solution vector set P converges to the suboptimal solution of the given problem.

Application example of optimization by genetic algorithm
As a reference, refer to the "Traveling salesman problem" in the above reference.
The indications of are described. Name of each of N cities is 1 ... N
It is represented by. Each element of the solution vector set P is a permutation of 1 ... N
Can be expressed as a vector. Also, the evaluation scale
Is the place where you have visited each city according to the order defined by this permutation.
It is the whole journey. For example, in the city A₁, A₂, ..., A_NIs allocated
And the city A₁-> A₂-> ...-> A_NPatrol in the order of
Then the corresponding solution vector is (A₁, A₂, ..., A _N)
It For example, simply A₁A₂A₃… A_NIt is written as. That
After that, by repeating the genetic algorithm,
A route similar to the short route can be obtained.

FIG. 67 shows a block diagram of a conventional sequential optimization processing apparatus using a genetic algorithm, which is executed according to the flow chart of FIG. 68 as described above. In FIG. 67, 101 is an initial solution set setting unit that sets an initial solution vector set to be improved, 102 is a genetic algorithm processing unit that actually estimates the optimal solution vector by genetic operation, and 103 is preset. It is an optimum solution output unit that outputs a solution vector with the highest degree of matching when the termination condition is satisfied. The genetic algorithm processing unit 102 calculates the suitability of each solution vector from the evaluation value acquisition unit 104 that acquires the evaluation value of each solution vector from a preset evaluation function and the evaluation value obtained by the evaluation value acquisition unit 104. Goodness-of-fit calculation unit 105, selection selection execution unit 670 that performs selection selection of a solution vector set based on the goodness-of-fit calculated by the goodness-of-fit calculation unit 105.
Crossover processing execution unit 202 that performs crossover processing between 1 and the solution vector
And a mutation processing execution unit 203 that performs the mutation processing of the solution vector. In the conventional optimization adjusting device using the genetic algorithm configured as described above, the evaluation value acquisition unit 104, the selection selection execution unit 6701, the crossover processing execution unit 202, and the mutation processing execution unit are executed according to the flowchart shown in FIG. The sub-optimal solution vector of the target optimization problem is estimated by repeatedly executing 203 and sequentially updating the solution vector set.

In the case of adjusting the characteristics of conventional hearing aids, designing products and interiors according to concepts that cannot be explicitly shown, and adjusting the acoustic and image characteristics according to individual tastes, the evaluation criteria are ideal. However, there is a problem that the evaluation of the adjustment result cannot be expressed quantitatively because it is unclear and the characteristics vary greatly depending on the individual. Therefore, it is impossible in principle to use a method of defining a clear evaluation function such as a neural network and performing adjustment so that the evaluation function is optimal for such a problem. For example, the characteristic adjustment of a hearing aid is performed in the following procedure. First, the test voice signal is divided into frequency bands, and test signals corresponding to the lower limit level and the upper limit level of the test voice are created in each frequency band. The test signals of the lower limit level and the upper limit level of the voice are input to the hearing aid, and the hearing aid user who is the user of the hearing aid hears the output test signals of these test signals from the hearing aid to confirm whether or not they can be comfortably heard. If the output test signal at the lower limit level of the sound in a certain frequency band is too small to be heard, an expert adjusts the frequency characteristic of the gain of the hearing aid so that the gain of the hearing aid in that frequency band increases.
On the contrary, if the output test signal of the upper limit level of the sound in a certain frequency band is too loud and uncomfortable, the expert adjusts the frequency characteristic of the gain of the hearing aid so that the gain of the hearing aid decreases in that frequency band. . The above is repeated until the user is satisfied. In this way, in order to fulfill the needs of users who are hearing impaired, it is the actual situation that specialists who adjust hearing aids adjust the frequency characteristics of gain of hearing aids based on the experience of each person. Yes, it takes a considerable amount of time and effort to find the characteristics that suit the hearing impaired. In addition, the parameter to be adjusted varies according to the usage environment and the preference of the hearing-impaired person, and one hearing-impaired person needs to prepare a plurality of hearing aids due to the influence of the parameter, so that the hearing aid must be adjusted repeatedly. there were. Each time, the hearing-impaired person had to go to an expert who could adjust the characteristics of the hearing aid.

In recent years, an interactive genetic algorithm for estimating an optimum solution based on the evaluation value of each solution vector determined by the user has been proposed in the methods of the above-described genetic algorithm. By using this interactive genetic algorithm, it is possible to make evaluations and decisions based only on human sensitivity and subjective evaluations. is there. For the conventional technique of the optimization adjustment method of the optimal solution vector by the interactive genetic algorithm, for example, see “Tracking Algorithm”.
Criminal Suspect Through "Face-Space" with Genetic Algorithm ("Tra
cking a Criminal Suspect through "Face-Space" with
a Genetic Algorithm ") (Caldwell, C. and Johnsto
n, VS: Proc.of 4th Int'l Conf. on Genetic Algorit
hms (ICGA'91)), "Evolutionary Art and Computers"
uters ") (Todd, S. and Latham, W .: Academic Press,
Harcourt, Brace, Jovanovich) and others.

The outline of the optimization adjusting method using the interactive genetic algorithm will be described below.

FIG. 70 is a flowchart showing the principle of optimizing and adjusting a solution vector by an interactive genetic algorithm. As shown in the case of the flowchart of FIG. 70, instead of the evaluation scale defined in advance by the functional expression in << step 2 >> of the optimization adjustment method by the genetic algorithm described above, the image represented by each solution vector is displayed to the user. There is a big difference in that information such as figures is presented and the goodness of each solution vector is evaluated.

As an application example of the optimal vector search by the interactive genetic algorithm, the above-mentioned reference describes an adaptation to "making a montage by a witness of a crime". A face is composed of constituent parts such as hair, eyes, nose, mouth, chin, and ears. Each part has a plurality of pattern images, and a facial photograph is created by combining the pattern images of these parts. The vector whose element is the pattern number of each part is the solution vector searched by the interactive genetic algorithm. Then, the user who is an eyewitness looks at the faces displayed on the screen from each solution vector and subjectively judges how similar he or she is to the person who witnessed them. We will create montage photographs that are supposed to express.

FIG. 69 shows a configuration diagram of a conventional optimization adjusting apparatus using an interactive genetic algorithm, which is executed according to the flow chart of FIG. 70 as described above. FIG. 69
In the above, an initial solution set setting unit 102, an interactive genetic algorithm processing unit 6901 that actually executes an optimal vector search by a genetic recombination operation, and an optimal solution output unit 104, and an interactive genetic algorithm processing unit 6901.
Is a presentation information unit 3207 that presents the information represented by each solution vector to the user, a user evaluation determination unit 3208 that determines the evaluation value of each solution vector based on the information presented by the presentation information unit, and a fitness calculation unit. 110, recombination operation unit 108
It is composed of The information presenting unit 3207 plays a role of presenting the information represented by each solution vector to the user, and the user evaluation determining unit 3208 asks the user to evaluate the goodness of each solution vector based on his / her own evaluation criteria and input the score. It corresponds to the part. Other than that, the description is omitted because it is almost the same as the configuration diagram of the optimization adjusting apparatus using the genetic algorithm in FIG.

[0015]

However, the genetic algorithm is an algorithm for performing a probabilistic and collective search, and uses only the information (evaluation value) of the current point in the search space. Therefore, in the above-described conventional sequential optimization processing device using a genetic algorithm, a global solution search is performed in comparison with a case where a solution search of a combinational optimum problem is performed using a neural network or the like that uses the characteristics of the curved surface of the evaluation function. Is excellent, but lacks the ability to search in the vicinity of the optimum solution because it does not use local information, and in the vicinity, there is a problem that it takes time to search the solution. Moreover, since the history information of the multi-point search performed so far is not used, there is a tendency to converge to the vicinity of the optimum solution while oscillating the search space to some extent.

Further, generally, in the optimization adjustment method by the interactive genetic algorithm, it is necessary to evaluate the preference of each solution vector by presenting a plurality of pieces of information represented by each solution vector to the user and making a comparative evaluation. There is. for that reason,
An optimization adjustment device using a conventional interactive genetic algorithm can be applied to adjustment of static data that can be spatially simultaneously presented such as characters and images without any problem.
When the user's evaluation target is time-series information, as in the case of the hearing aid characteristic adjustment problem described above, it is difficult for the user to distinguish the differences in the information even if multiple pieces of each time-series information are presented as they are. It is very difficult to evaluate a vector. In addition, the burden on the user to make adjustments increases as time goes by, and his / her taste changes greatly. Nevertheless, the conventional adjustment device using the interactive genetic algorithm does not take any measures against these problems.

The present invention is to solve the above-mentioned problems.
Optimization that performs fast and efficient estimation of optimal solutions without impairing the global search ability originally possessed by the genetic algorithm by adding ingenuity to local search ability to the conventional genetic algorithm method It is an object to provide an adjusting device.

Further, an optimization adjusting device capable of efficiently applying the interactive genetic algorithm to not only the adjustment of characters and images but also the adjustment of time series signals, and the efficiency of the process of optimizing the solution vector are aimed at. Alternatively, an interactive genetic algorithm that reduces the burden on the user by modeling the user's adjustment process from the obtained history of the user's adjustment process and automatically adjusting the solution vector using it It is an object of the present invention to provide an optimization adjusting method and an optimization adjusting device for a device used.

[0019]

In order to achieve the above object, a first optimization adjusting method and an optimization adjusting apparatus according to the present invention are configured so that each solution vector in a solution vector set is centered on that vector. The initial solution vector group is set again in the vicinity of and the recombination operation is performed a certain number of times to find a more optimal solution vector group. Then, the obtained plurality of solution vector groups are integrated into one large set, and a recombination operation is performed to update the optimum solution vector again. As described above, the second sequential optimization processing device locally updates the solution vector several times by using the recombination operation and then updates the global solution vector again by the recombination operation to obtain the optimum solution vector. It is an estimate.

In order to achieve the above object, the second aspect of the present invention
The optimization adjusting method and the optimization adjusting device of (1) select a solution vector having a higher degree of conformity as compared with each solution vector in the solution vector set and a vector group randomly extracted from the neighboring space as a solution vector belonging to the solution vector set. , Is designed to estimate the optimal solution vector by a recombination operation on the resulting solution vector set.

In order to achieve the above object, the third aspect of the present invention
The optimization adjustment method and the optimization adjustment apparatus of (1) extract a neighborhood vector group of the selected solution vector and perform a genetic recombination operation only within the neighborhood vector group to search for an optimal solution. Is.

In order to achieve the above object, the fourth aspect of the present invention
The optimization adjusting method and the optimization adjusting apparatus of (1) divide each solution vector neighborhood in the solution vector set into a plurality of areas and randomly select a plurality of solution vector groups from the area having the largest average goodness of fit. A new solution vector set is generated by a recombination operation on the solution vector in the vector set. Then, from these, the solutions are selected in descending order of suitability and integrated into one solution vector set. This sequential optimization processing device updates each solution vector in the solution vector set by recombination operation in parallel with each solution vector. It also updates the solution vector in the vicinity.

In order to achieve the above object, the fifth aspect of the present invention
The optimization adjustment method and the optimization adjustment device are optimized by dividing the solution vector set into multiple groups based on the arithmetic mean and standard deviation of the goodness of fit and performing the recombination operation on the solution vectors in the group. In order to achieve the above object, which is to search for a solution, the sixth optimization adjusting method and the optimization adjusting apparatus of the present invention provide an arithmetic mean and a standard deviation of the goodness of fit of each solution vector, Based on the maximum and minimum goodness of fit, it is determined whether the solution vector set is divided. When it is judged that it is better to divide, the solution vector set is divided into multiple groups and the recombining operation is performed on the solution vectors in that group. If not so, the whole solution vector set is reconstructed. Do the operation. The seventh optimization method and its adjusting apparatus of the present invention estimate the optimum solution by the above processing.

In order to achieve the above object, the seventh aspect of the present invention
The optimization adjustment method and the optimization adjustment device of (1) first roughly estimate the distribution area of the optimum solution vector by globally dividing the space and updating the solution vector, and then gradually converge the solution vector. The optimal solution vector estimation is performed by tightening the step convergence criterion for determining and updating the solution vector locally by subdividing the update region.

The first optimization adjusting method and the optimization adjusting apparatus are based on the goodness of fit of each solution vector belonging to the solution vector set, and have a centroid of the solution vector having a goodness of fit which is equal to or higher than a predetermined reference goodness of fit. The update direction candidate vector is calculated based on the vector and the centroid vector obtained in the past update, and if the direction matches the recorded past update direction candidate vector, the direction in which the optimal solution vector is distributed. Determines that this vector is in the direction of that vector, considers this vector as the update direction vector again, updates a plurality of solution vectors along that vector, and the remaining solution vectors of the solution vectors in the solution vector set before updating. It is updated by a recombination operation based on a genetic operation. When the update direction vector cannot be obtained, the entire solution vector set is targeted for the recombination operation based on the genetic operation. As described above, the first optimization adjusting method and the adjusting apparatus estimate and use the distribution direction of the optimum solution vector based on the history of updating the solution vector in the past, so that the optimum solution vector is fast and efficient. Can be estimated.

The second optimization adjusting method and the optimization adjusting apparatus first set an appropriate initial solution vector set, and then set the initial solution vector group again in a limited area around it according to each solution vector. A local update of the solution vector is realized by a recombination operation based on the degree. After that, the plurality of solution vector groups obtained by the local update are integrated into one large set, and the optimal solution vector is globally estimated by recombining the solution vectors according to the goodness of fit. In this way, after the local solution update by the recombination operation, the solution vector on the global scale is updated again by the recombination operation, thereby compensating for the local solution update ability that is lacking in the conventional genetic algorithm. You can

The third optimization adjusting method and the optimization adjusting apparatus first set the initial solution vector set appropriately and then extract a plurality of solution vectors by limiting the surrounding area for each solution vector. Among the extracted solution vectors in the solution vector group, a solution vector having a high fitness is selected and the solution vector set is reset again. Then, a solution vector recombination operation is performed on the reset solution vector set based on the goodness of fit of each vector, so that the third optimization adjusting method and its adjusting device can obtain a local solution. The update capability can be enhanced and the optimal solution vector can be estimated at high speed.

The fourth optimization adjusting method and the optimization adjusting apparatus first set an appropriate initial solution vector set, and then determine the goodness of fit for each solution vector belonging to the solution vector set. Then, based on the obtained goodness of fit, a representative solution vector is selected and its neighboring vector group is extracted. The solution vector rearrangement operation is performed in the group rearrangement operation unit for the solution vectors in this neighborhood vector group. After that, the newly created solution vector group is again in charge of the recombination operation by the mutation processing. By taking such a procedure, it is possible to avoid that the influence of a solution vector with a high degree of goodness of conformity has a large influence on the entire solution immediately, and it is possible to efficiently estimate the optimal solution. is there.

The fifth optimization adjusting method and the optimization adjusting apparatus first set the initial solution vector set appropriately, then find the goodness of fit of each solution vector belonging to the solution vector set, and according to this goodness of fit, the solution vector set Perform the recombination operation of the solution vector in. On the other hand, a plurality of solution vectors are arbitrarily extracted from the solution vector set, and these neighboring spaces are divided into a plurality of regions. Multiple solution vectors are extracted for each region,
Based on the average evaluation value and the average goodness of fit for each area, a plurality of solution vector groups are arbitrarily selected from the area having the highest average goodness of fit. A solution vector having a high degree of conformity is sequentially selected from the selected solution vector group and the solution vector group generated by the recombination operation to generate a solution vector set which is the next processing target. By repeating such a procedure, it is possible to enhance the ability to update the solution locally while utilizing the efficient global solution updating ability of the genetic algorithm, and to efficiently estimate the optimal solution vector. can do.

The sixth optimization adjusting method and the optimization adjusting apparatus first set the initial solution vector set appropriately, and then find the goodness of fit corresponding to each solution vector belonging to the solution vector set. Based on the arithmetic mean and standard deviation of the goodness of fit, the entire solution vector set is divided into a plurality of groups. The solution vector in each group is selected, and the solution vector is updated by selection of the solution vector and recombination operation. Then, the recombination operation is performed again on the newly created solution vector by the mutation process. By taking such a procedure, it is possible to avoid the influence of a solution vector having a high degree of conformity having a great influence on the entire solution immediately, and it is possible to efficiently estimate the optimum solution.

The seventh optimization adjusting method and the optimization adjusting apparatus first set the initial solution vector set appropriately, and then find the goodness of fit corresponding to each solution vector belonging to the solution vector set. The arithmetic mean and standard deviation of this goodness of fit, and the maximum
Based on the minimum goodness of fit, it is determined whether to target the entire solution set or to divide into a plurality of groups and perform the recombination operation. If it is determined to be divided, the entire solution set is divided into a plurality of groups as in the sixth optimization adjusting method and the optimization adjusting apparatus. If not so, the entire solution set is set as the target of the recombination operation. To do. After that, the solution vector is selected and the recombination operation is performed on the obtained recombination processing target. As described above, it is possible to determine whether there is a solution vector having a high degree of goodness of conformity based on the distribution of goodness of fit, and to limit the recombination operation target so as to reduce the effect of the solution vector. An estimate can be made.

The eighth optimization adjusting method and the optimization adjusting apparatus first set the initial solution vector set appropriately and then derive the goodness of fit of each solution vector belonging to the solution vector set.
Based on this goodness of fit, the genetic recombination operation is used to select and recombine the solution vector. At that time, the solution vector in the set update region is satisfied until the set step convergence criterion is satisfied. Execute the recombination operation for. Then, when the gradual convergence criterion is satisfied, the gradual convergence criterion is further tightened and the update area is narrowed, and the processes such as the fitness calculation and the recombination operation are repeated. By taking such a procedure, a global search is first performed and then a local update is gradually performed, so that an efficient optimal solution estimation can be realized.

[0033]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. 1 is a block diagram of an optimization adjusting apparatus in a reference example related to the present invention, FIG. 2 is a block diagram of a recombination operation unit which is a main part of the optimization adjusting apparatus in a reference example, and FIG. 10 is a first example of the present invention. FIG. 13 is a block diagram of an optimization adjusting device according to an embodiment, FIG. 13 is a block diagram of an optimization adjusting device according to a second embodiment of the present invention,
FIG. 16 is a block diagram of an optimization adjusting device according to the third embodiment of the present invention, FIG. 19 is a block diagram of an optimization adjusting device according to the fourth embodiment of the present invention, and FIG. 22 is a fifth view of the present invention. 25 is a block diagram of an optimization adjusting apparatus according to the embodiment of the present invention, FIG. 25 is a block diagram of an optimization adjusting apparatus according to the sixth embodiment of the present invention, and FIG. 28 is an optimization adjustment according to the seventh embodiment of the present invention. 32 is a block diagram of an optimization adjusting device in a first example related to the present invention, FIG. 34 is a block diagram of an optimization adjusting device in a second example related to the present invention, and FIG. 38 is a block diagram of an optimization adjusting apparatus in a third example related to the present invention.
FIG. 40 is a block diagram of an optimization adjusting device in a fourth example related to the present invention, FIG. 40 is a block diagram of an optimization adjusting device in a fifth example related to the present invention, and FIG. 43 is a sixth diagram related to the present invention. 46 is a block diagram of the optimization adjusting apparatus in the example of FIG. 46, FIG. 46 is a block diagram of the optimization adjusting apparatus in the seventh example related to the present invention, and FIG. 49 is a diagram of the optimization adjusting apparatus in the eighth example related to the present invention. FIG. 51 is a block diagram of an optimization adjusting apparatus in a ninth example related to the present invention, FIG. 53 is a block diagram of an optimization adjusting apparatus in a tenth example related to the present invention, and FIG. 57 is a block diagram of an optimization adjusting device in an eleventh example related to FIG.
59 is a block diagram of an optimization adjusting apparatus in a twelfth example related to the present invention, FIG. 59 is a block diagram of an optimization adjusting apparatus in a thirteenth example related to the present invention, and FIG. 62 is a fourteenth related to the present invention. FIG. 65 is a block diagram of the optimization adjusting apparatus in the example of FIG. 65, and FIG. 65 is a block diagram of the optimization adjusting apparatus in the fifteenth example related to the present invention.

Further, FIG. 3 is a flow chart showing the processing steps of the optimization adjusting method in the reference example of the present invention, and FIG. 5 shows the steps of the recombination operation processing which is the main processing of the optimization adjusting method in the reference example of the present invention. Flow chart diagram, FIG.
FIG. 17 is a flow chart showing the processing steps of the optimization adjusting method according to the first embodiment of the present invention. FIG. 14 is a flow chart showing the processing steps of the optimization adjusting method according to the second embodiment of the present invention. 2 is a flow chart showing the processing steps of the optimization adjusting method according to the third embodiment of the present invention. FIG. 20 is a flow chart showing the processing steps of the optimization adjusting method according to the fourth embodiment of the present invention.
3 is a flowchart showing the processing steps of the optimization adjusting method according to the fifth embodiment of the present invention, and FIG. 26 is a flowchart showing the overall processing steps of the optimization adjusting method according to the sixth embodiment of the present invention. FIG. 27 is a flow chart showing the process of Process 2 of the optimization adjusting method according to the sixth embodiment of the present invention, and FIG. 29 shows the process of the optimization adjusting method according to the seventh embodiment of the present invention. FIG. 33 is a flowchart showing the processing steps of the optimization adjusting method in the first example related to the present invention.
5 is a flowchart showing the processing steps of the optimization adjusting method in the second example related to the present invention, and FIG. 37 is a flowchart showing the processing steps of the optimization adjusting method in the third example related to the present invention. 39 is the fourth related to the present invention
41 is a flow chart showing the processing steps of the optimization adjusting method in the above example, FIG. 41 is a flow chart showing the processing steps of the optimization adjusting method in the fifth example relating to the present invention, and FIG. 42 is the fifth step relating to the present invention. 44 is a flowchart showing the individual adjustment processing steps of the optimization adjustment method in the example of FIG.
Is a flow chart showing the processing steps of the optimization adjusting method in the sixth example related to the present invention, and FIG. 45 is a flow chart showing the steps of the common model updating processing of the optimization adjusting method in the sixth example related to the present invention. FIG. 47 is a flow chart showing the processing steps of the optimization adjusting method in the seventh example related to the present invention, and FIG. 48 is a continuation of the processing steps of the optimization adjusting method in the seventh example related to the present invention. 50 is a flowchart showing the processing steps of the optimization adjusting method in the eighth example related to the present invention. FIG. 52 is a flowchart showing the processing steps of the optimization adjusting method in the ninth example related to the present invention. 54 is a flow chart showing
FIG. 56 is a flow chart showing the processing steps of the optimization adjusting method in the tenth example related to the present invention, FIG. 56 is a flow chart showing the processing steps of the optimization adjusting method in the eleventh example related to the present invention, and FIG. Twelfth related to the present invention
FIG. 6 is a flowchart showing the processing steps of the optimization adjusting method in the above example, FIG. 60 is a flowchart showing the processing steps of the optimization adjusting method in the thirteenth example related to the present invention, and FIG.
1 is a flow chart showing the process of the individual adjustment process 2 of the optimization adjusting method in the thirteenth example related to the present invention, and FIG. 63 shows the process of the optimization adjusting method in the fourteenth example related to the present invention. 64 is a flowchart showing the steps of the common model updating process 2 of the optimization adjusting method in the fourteenth example related to the present invention.
6 is a flow chart showing the processing steps of the optimization adjusting method in the fifteenth example related to the present invention. In each of the block diagrams, the same parts are designated by the same reference numerals.

The reference examples of the present invention and the first to seventh embodiments relate to an optimization adjusting method and an optimization adjusting apparatus using a genetic algorithm. The first to fifteenth examples are user's. The present invention relates to an optimization adjusting method and an optimization adjusting device using an interactive genetic algorithm for estimating an optimal solution based on the evaluation of.

The optimization adjusting method and adjusting apparatus in the reference example of the present invention will be described below. In this reference example, the optimal solution vector is efficiently estimated by predicting and utilizing the direction in which the solution vector having a high degree of conformity is distributed based on the history of past update of the solution vector.

In FIG. 1, 101 is an initial solution set setting for setting an initial solution vector set P = [p _k ] (k = 1, ..., n) to be updated by a predetermined procedure or an instruction from the outside. Part, 1
Reference numeral 02 denotes a genetic algorithm processing unit that actually estimates an optimal solution vector by a genetic recombination operation, and reference numeral 103 denotes a solution vector with the highest degree of fitness from the latest solution vector set when a preset end condition is satisfied. It is an optimum solution output unit that outputs as an optimum solution vector. The genetic algorithm processing unit 102 includes an evaluation value acquisition unit 104 that calculates an evaluation value E _k of each solution vector for the solution vector set P according to a preset evaluation function, and an evaluation value acquisition unit 10.
4. The fitness calculation unit 105 that calculates the fitness f _k of each solution vector for the problem based on the evaluation value E _k obtained in step 4, and the solution having a fitness higher than the predetermined standard fitness f ^th. An update direction determination unit 106 that selects a vector group P ′ and determines whether or not an update direction vector in which a centroid vector g _l (l is the number of iterations) and a solution vector having a higher degree of fitness than the past centroid vector can be estimated. Update direction determination unit 10
When the update direction vector is estimated in 6, a plurality of solution vector updates are executed along the direction, and the direction application update unit 107 that sets the number of solution vectors generated by the recombination operation and the update direction determination unit 106. Alternatively, in response to the result of the direction application updating unit 107, a recombination operation unit 108 that generates a solution vector by a recombination operation based on the genetic operation of the solution vector in the original solution vector set by the set number of solution vectors is performed. Composed. Further, the update direction determination unit 106 selects a solution vector group P ′ having a goodness of fit, which is higher than a predetermined standard goodness of fit f ^th, with respect to the goodness of fit obtained by the goodness of fit calculation unit 105, and selects its centroid vector _gl Centroid estimation unit 109 and centroid estimation unit 10 that calculate
Obtains a differential vector △ g _l-1 of the resulting g _l past centroid vectors g _l-1 at 9, update direction candidate vector v _'l-1 = △ g recording to as _l-1 update direction candidate recording unit 110 and the degree of coincidence of the _v'l and _v'l-1 directions stored in the update direction candidate recording unit 110 are evaluated, and if it is determined that they match, the update direction vector v = v ' _The process is passed to the direction application updating unit 107 as if _l has been obtained, and if they do not match, the entire solution vector in the solution vector set is targeted for recombination operation or the process is transferred to the recombination operation unit 108.
It is composed of 1. The direction application update unit 107 updates the gravity center vector obtained by the gravity center estimation unit 109 to the update direction vector v.
Center of gravity moving unit 1 that moves according to the reference moving distance length
12 and a centroid surrounding solution vector generation unit 113 that generates a solution vector of a preset number of individuals around the centroid vector newly obtained by the centroid moving unit 112.
As shown in FIG. 2, the recombination operation unit 108 selects a solution vector from the solution vector set P by using the goodness of fit f _k, and selects a solution vector from the candidate selection unit 201 and the solution vector set obtained by the candidate selection unit 201. The crossover processing execution unit 202 that executes the crossover processing by the crossover processing, and the mutation processing execution unit 203 that executes the mutation processing on the solution vector set obtained by the crossover processing execution unit 202. Furthermore, the candidate selection unit 20
1 is a selection range derivation unit 204 for deriving a selection probability h _k and a selection range I _k when a certain solution vector is selected from the solution vector set, and a set R = of uniform random numbers r _k in [0,1] (r ₁ , r ₂ , ..., r _n )
And a solution vector extraction unit 206 that extracts a solution vector to be selected from a solution vector set based on the result of the random number generation unit 205.

In the reference example of the present invention configured as described above
The operation of the optimizing adjustment device will be described. That
In this reference example, as a specific problem to be dealt with, as shown in FIG.
M-dimensional function w (x₁, x₂, ..., x_m) Maximum value estimation problem
Get out. As shown in FIG. 6, for example, the maximum of the m-dimensional function w
In the value estimation problem, <Condition 1> According to a predetermined procedure or an external instruction,
n m-dimensional vectors q_i(x ₁ ⁱ, x₂ ⁱ, ..., x_m ⁱ) (i = 1, ..., n)
Create a set Q consisting of <Condition 2> Each q_iFunction value w (q at (i = 1, ..., n)_i)
Therefore, based on that value, the m-dimensional vector that maximizes the function value w (q)
Estimate the tor q. <Condition 3> Each element x_kThe absolute value of | x_k| ≤ c_kTo be satisfied. <Condition 4> Also each element x_kIs the precision up to a digit after the decimal point
I do not have it. Such a procedure is taken. Thus the m-dimensional function w
For the maximum value estimation problem, the flow of FIG. 3, FIG. 4, and FIG.
-Operation of the first embodiment of the present invention based on a chart
Will be described.

First, the initial solution set setting unit 101 sets an initial set of solution vectors P = [p _k ] (k = 1, ..., n) according to a predetermined procedure or an instruction from the outside. Various methods can be considered as this setting method. Here, it is assumed that the element x ⁱ _j (j = 1, ..., m) of the m-dimensional vector q _{i to be} obtained is a real value having an accuracy of one digit below the decimal point, and the value is shown in FIG. It is converted into a bit string code of length Blen which is associated as described above. Then, it is assumed that the solution vector p _k (k = 1, ..., n) is represented by arranging the elements obtained by converting each element x ⁱ _j of the vector q _i into a bit string code in order. In addition,
Here, the solution vector is expressed by the fixed-length bit string code in this way, but the expression method of the solution vector is not limited to this method, and it is conceivable that the solution vector may be directly handled as it has elements as real elements.

According to the expression method using the above-mentioned bit string code, the l-th solution vector expressed in the form of a fixed length bit string code based on the uniform random number r in [0,1) and (Equation 1). The initial solution vector set P = [p _k ] (k = 1, ..., n) is set by obtaining the bits (l = 1, .., Blen × m) of. Here, in (Equation 1), u _l represents the value at the l-th bit from the lower order.

[0041]

[Equation 1]

Optimal solution vector estimation is started from this initial solution vector set P. At the same time, the center of gravity estimation unit 10
The reference conformance f ^th used in 9 and the reference moving distance length used in the center-of-gravity moving unit 112 are set.

The evaluation value acquisition unit 104 is a part for obtaining an evaluation value E _k for each solution vector p _k (k = 1, ..., N) as described above, and for example, an evaluation function such as (Equation 2) Ask by.

[0044]

[Equation 2]

Here, q _k is a coordinate vector when the solution vector p _k is returned to the original m-dimensional space, and w ^t _min is a set P
In the function value w (q _k ) when each solution vector p _k in is returned to the m-dimensional space coordinates, the number of update times is l Represents a value. As is clear from (Equation 2), it represents a value obtained by normalizing the difference from the minimum value of the function values in the solution vector set obtained so far as the entire set. In the present embodiment, this can be regarded as a maximization problem that maximizes the evaluation value.

In the fitness calculation unit 105, the evaluation value acquisition unit 1
From the evaluation value obtained in 04, the goodness of fit is calculated in order to see the suitability of each solution vector. Various functions can be considered as a function for deriving the goodness of fit f _k . Here, by setting f _k = E _k , the higher the evaluation value, the higher the goodness of fit, and it becomes possible to handle the maximum value estimation problem. .

Next, the operation of the update direction judging section 106 will be described. First, in the centroid estimation unit 109, the fitness f _{k of} the solution vector p _k (k = 1, ..., N) obtained by the fitness calculation unit 105 is compared with the reference fitness f ^th , and its value f ^th Select a solution vector group P '= [p _j ^l ] (j = 1, .., p # num) having a larger goodness of fit. p # num is the number of solution vectors belonging to P ',
At this time, the center-of-gravity vector _gl is obtained as in (Equation 3). Here, l represents the number of repetitions of the updating process in the genetic algorithm unit 102.

[0048]

[Equation 3]

The update direction candidate recording unit 110 is obtained at 109.
Center of gravity vector g _lAnd at the (l-1) th time
Centroid vector g_l-1Difference vector of Δg_l-1How to update
Candidate vector v^' _l-1= △ g_l-1To record as. Update direction
In the acquisition unit 111, the update direction candidate recording unit 110 records
Updating direction candidate vector v^' _lAnd v^' _l-1Degree of coincidence
Is evaluated, the angle θ formed by the two vectors is calculated,
Matching angle θ that is given in advance^thCompared to
If it is smaller than that value, two update direction candidate vectors v^' _l
And v^' _l-1Are judged to face the same direction. θ is
It can be obtained as in (Equation 4). Note that cos^-1Is
Inverse cosine function (inverse cosine function), (x, y) is the vector x
The dot product of y, || x || is the norm (magnitude) of vector x
You

[0050]

[Equation 4]

[0051] If If v _^'l and ^v' _l-1 is judged to be the same direction, and v = v _^'l as updating the direction vector v is obtained. Then, in this case, the processing shifts to the direction application updating unit 107. On the other hand, when it is determined that v ^' _l and v ^' _l-1 are not in the same direction, all n solution vectors p _k in the solution vector set P are solved by the recombination operation based on the genetic operation in the recombination operation unit 108. The vector is updated. FIG. 8 schematically shows the above situation.

First, the processing when the update direction vector v is obtained will be described. In this case, the direction application update unit 1
The processing moves to 07, but the next center of gravity moving unit 1
12 and the centroid surrounding solution vector generation unit 113 are processed in this order. The center-of-gravity moving unit 112 performs a process of moving the center-of-gravity vector _gl calculated by the center-of-gravity estimating unit 109 in the direction indicated by the update direction vector v. For simplicity here,
A method of estimating a new center-of-gravity vector g _{l + 1} by moving the translation distance length set in advance is adopted. However, a method is also conceivable in which a point having the maximum evaluation value is obtained by using a line search method such as the PATAN method in the direction of the update direction vector v, and the point is set again as the center of gravity vector g _{l + 1} . The center-of-gravity surrounding solution vector extraction unit 113 calculates 11 as in (Equation 5).
Radius around the new center of gravity vector g _{l + 1} obtained in 2
A # s points arbitrarily included in the area inside the sphere in the multidimensional space of a # len are extracted and used as an element of a new solution vector set.

[0053]

[Equation 5]

Although a # s and a # len are considered here as preset constant values, a method for dynamically changing these values,
This value is converted to a bit string as shown in Fig. 7, and the solution vector
In addition to p _k , a method of defining it as a solution vector again can be considered. Further, in 113, (n-a # s) is set as a new number of solution vectors generated by the rearrangement operation performed in the next rearrangement operation unit 108.

Next, the operation of the recombination operation section 108 will be described. Here, in the solution vector set P at the number of iterations l, the number of solution vectors set by the update direction acquisition section 111 or the centroid surrounding solution vector extraction section 113 is set. A new solution vector is generated by recombining the solution vector of. The processing procedure from the candidate selecting unit 201 to the solution vector extracting unit 206 will be described below. First, the candidate selection unit 201 executes selection and selection of solution vectors. In this case, as shown in FIG. 9, a roulette selection method that selects a solution vector with a probability proportional to the goodness of fit is used.

(Roulette selection method) (i) Goodness of fit f of each solution vector p _k (k = 1, ..., n) belonging to the set P
_k , the total sum f of goodness of fit of all solution vectors is obtained. (ii) The selection probability h _k that p _k is selected as the parent that creates the next-generation solution vector is calculated as in (Equation 6).

[0057]

[Equation 6]

To assign this probability to the solution vector, for example, the following method can be considered. (iii) The selection range I _k of each solution vector is assigned to the section in [0,1) using (Equation 7) and (Equation 8) as follows. That is,

[0059]

[Equation 7]

When the [0060], selected range I _k of p _k is

[0061]

[Equation 8]

It is defined as follows. Here, a set of uniform random numbers r _k R = (r ₁ , r ₂ , ..., R _n ) is generated in [0, 1]. r _j ∈ I
A set of num _j = k satisfying _k (j, k = 1, ..., n) Num = (num ₁ , nu
m ₂ , ..., num _n )
A set of solution vectors will be selected.

The solution vector p _{k in} the current solution vector group P is selected by such a roulette selection method. First, according to (Equation 6) to (Equation 8), the selection range deriving unit 204 obtains the probability h _k that each solution vector is selected and its selection range I _k . Then, the random number generator 205 changes from 0 to 1
Generates n uniform random numbers r between. The set R of random numbers obtained by the random number generation unit 205 and the selection range I _k obtained by the selection range derivation unit 204 are sent to the solution vector extraction unit 206 and r _j ε
A set Num of num _j = _k that satisfies I _k is obtained. As a result, the solution vector extraction unit 206 outputs a new solution vector group P composed of solution vectors specified by Num. The crossover process execution unit 202 performs the crossover process on the new solution vector group P obtained by the candidate selection unit 201. As described above, there are various methods for the crossover processing, but in this reference example, the one-point crossover processing or the two-point crossover processing as shown in FIG. 72 is used. Furthermore, the mutation processing unit 2
03 performs a mutation process for performing bit inversion on each new solution vector group obtained through the crossover process execution unit 202 with a low probability that each bit forming the solution vector has a low probability. At that time, we tried to maintain the diversity of solution vectors by changing the probability of mutation in half and the other half of the solution vector group. In addition, here, the solution vector to be obtained is converted into a bit string code and treated, but as described above, it is also possible to arrange the real-valued elements of the coordinate vector in the multidimensional space as they are and treat them as a solution vector. In this case, the crossover process is similar to the case of the bit string code, and the mutation process is a random number given within a certain range to the gene (element of the coordinate vector in the multidimensional space) selected with a certain low probability. It is realized by adding.

Finally, the termination condition of the genetic algorithm unit 102 will be described. Various end conditions for the solution vector updating process by genetic recombination of 102 may be considered depending on the problem to be applied. Here, << end condition 1 >> the goodness of fit f obtained by the goodness-of-fit calculation unit 105 _k
Whether the maximum value f ^{max of} is greater than the convergence determination fitness f ^end << End condition 2 >> Repeat count l is end update Repeat count g
Two ^end conditions are set to determine whether or not #num ^end is exceeded. Then, when neither condition is satisfied, the process returns to the evaluation value acquisition unit 104 again. The optimal solution vector is estimated by repeatedly executing the above-described processing steps until either end condition 1 or 2 is satisfied.
In this way, focusing on the movement vector of the center of gravity vector of the group having a high degree of conformity in the solution vector set, the solution vector group in which the distribution direction in which the solution vector having a high degree of fitness exists is estimated is updated. At the same time, by optimizing the solution vector by the recombination operation on the original solution vector set at the same time, it is possible to estimate a fast and efficient optimal solution using the history of past solution vector update. Of.

The optimization adjusting method and the optimization adjusting apparatus according to the first embodiment of the present invention will be described below with reference to the drawings. FIG. 10 shows the configuration of the optimization adjusting apparatus according to the first embodiment of the present invention. In this embodiment, the solution vector is locally updated by using the recombination operation, and then the global solution vector is updated again by using the recombination operation to estimate the optimum solution vector. ..

In FIG. 10, reference numeral 1001 denotes a new initial solution vector group P around each vector p _k (k = 1, ..., n) in the initial solution vector set P set by the initial solution set setting unit 101.
Set ^k, and the solution vector p that belongs to the initial solution vector group
^k _j (j = 1, .., n ^k , k = 1, .., n) is a goodness of fit by repeating the recombination operation of the solution vector a preset number of times based on the goodness of fit f ^k j for the problem. , A local update unit for extracting a solution vector group P ^k having a high value, and a plurality of solution vector groups P ^k obtained by the local update unit 1001 are integrated into one set P ^* to obtain a solution vector p ^* _k in the set. It is a global updating unit that performs a recombination operation of (k = 1, ..., n ^* ). The local update unit 1001 determines a range when a local solution vector is updated from the solution vector p _k in the solution vector set P, and a local update setting unit 1003.
Initial solution vector group again within the range specified by 1003
A vector group setting unit 1004 that generates P ^k and a solution vector
Evaluation value E ^k _j for the target problem of p ^k _j (j = 1, .., n ^k , k = 1, .., n)
And an evaluation value acquisition unit 104 for acquiring
The fitness calculation unit 105 that calculates the fitness f ^k _j of each solution vector from the evaluation value obtained in step S, and the recombining of the solution vectors in the solution vector group based on the fitness obtained by the fitness calculation unit 105. A local recombination operation unit 1005 that executes an operation so as to be included within the range set by the local update setting unit 1003.
And a local update end determination unit 1006 that determines whether or not a series of local update processes satisfies a predetermined number of iterations. On the other hand, the global update unit 1002
Is the solution vector group P ^k (k =
1, ..., n) into one set P ^* = [p ^* _i ] (i = 1, ..., n ^* ), and a solution vector set P obtained in 1007 ^The global recombination operation unit 1008 updates the solution vector by the recombination operation for ^* . Further, both the local recombination operation unit 1005 and the global recombination operation unit 1008 are configured by a candidate selection unit 201, a crossover processing execution unit 202, and a mutation processing execution unit 203. It is composed of a unit 205 and a solution vector extraction unit 206.

The operation of the optimizing adjustment apparatus according to the first embodiment of the present invention constructed as above will be described with reference to the flowchart of FIG. Here, as a specific problem to be dealt with, the maximum value estimation problem of the multidimensional function w will be taken up as in the reference example of the present invention. Similarly, each element forming the coordinate vector in the multidimensional space is also expressed as a bit string code by the method shown in FIG. Shall be expressed.

First, in the initial solution set setting unit 101, an initial set of solution vectors P = [p _k ] (k = 1, ..., n) is set by a predetermined procedure or an instruction from the outside. , The local range radius b # used by the local update setting unit 1003
len and the local update end count c # l ^th used in the local update end determination unit 1006 are set. Local update setting unit 1003
Then, a spherical region in the multidimensional space within the local range radius b # len is set as in (Equation 9) with p _k in the initial set P as the center.

[0069]

[Equation 9]

The c # l ^th is fixed here, but it can be changed dynamically. For example, the value may be converted into a bit string and added to the solution vector p _k . The vector group initial setting unit 1004 arbitrarily extracts a plurality of solution vectors from the range set in 1003 and newly sets the initial solution vector group P ^k . Here, n ^k number of solutions, including the solution vector p _k became center vector p ^k _j (j =
1, ..., n ^k ). In this embodiment, n ^k = n.
The evaluation value acquisition unit 104 obtains the evaluation value E ^k _j of each solution vector according to (Equation 4), and the fitness calculation unit 105 calculates the fitness f ^k _j from the evaluation value E ^k _j . The local recombination operation unit 1005
Based on the goodness of fit in 105, the recombining operation of the solution vector is performed so that the solution vector is included in the area represented by (Equation 9).
This processing is the same as that of the recombination operation unit 108 of the optimization adjustment apparatus according to the first exemplary embodiment of the present invention, and therefore will be omitted. After updating the solution vector group P ^k in 1005,
The local update end determination unit 1006 compares the local update count c # l with the local update end count c # l ^th to determine whether or not c # l ≧ c # l ^th is satisfied. If satisfied, the local update process ends. If not satisfied, 1 is added to c # l, and the processing of the evaluation value acquisition unit 105 to the local update end determination unit 1006 is repeated. In this way, a recombination operation based on a genetic operation is performed around the initially set initial solution vector set to search for a more suitable solution vector. Note that n ^k in 1003 is
Although n ^k = n is set, it is possible to make it variable with respect to the solution vector p ^k in the original solution vector set P. It is also possible to dynamically change the local update count c # l.

Next, the processing in the global updating unit 1002
explain about. First, each solution vector extracted in 1001
Group P^k(k = 1, ..., n) is set by the set integration unit 1007 as one set P.
^*= [p ^* _i] (i = 1, ..., n^*). At this time, all solutions
Cuttle group P^kSort the solution vectors in
And n from highest to lowest^*Choose the solution vector
It is possible to generate a union set P. Global recombination operation
Sakube 1008 uses the solution vector set P obtained in 1007.^*
Update the solution vector by the recombination operation for
U In that case, this process is the optimization adjustment device in the reference example.
Since it is the same as the recombination operation unit 108, the description thereof will be omitted. Figure
12 is a diagram schematically showing a series of states.

Finally, the termination condition will be described. As the termination condition, the convergence determination suitability f ^end and the termination update repetition number g # num ^end are 2 as in the optimization adjustment apparatus in the reference example.
Value comparison conditions are set. Specifically, in the << end condition 1 >> set integration unit 1007, the maximum value f ^{max of the} goodness of fit of the solution vector p ^k i in each solution vector group P ^k
Is greater than the convergence determination fitness f ^end << End condition 2 >> Local update unit 1001 to global update unit 10
Two termination conditions are set to determine whether the number of times l of the series of updates in 02 exceeds the number of times of repeated end updates g # num ^end . Then, when neither condition is satisfied, the process returns to the local update setting unit 1003. When satisfied, the process moves to the optimum solution output unit 103, and the solution vector with the highest degree of matching in the latest solution vector set is output as the optimum solution vector. The second optimization adjustment apparatus in the present embodiment estimates the optimum solution vector by repeatedly executing the above-described processing steps until either end condition 1 or 2 is satisfied. In this way, the solution vector is locally updated using the recombination operation several times, and then the global solution vector is updated again by the recombination operation. It is possible to reinforce the local updating ability of the solution vector without impairing the above, and to realize the estimation of the optimal solution at high speed.

The optimization adjusting method and the optimization adjusting apparatus according to the second embodiment of the present invention will be described below. FIG. 13 shows the configuration of the optimization adjusting apparatus according to the second embodiment of the present invention. In the present embodiment, each solution vector in the solution vector set is compared with a vector group randomly extracted from the neighborhood space of the solution vector, and a vector group having a high degree of conformity is re-assigned to a solution vector set that is to be updated. It is characterized in that it is selected as and is targeted for recombination operation.

In FIG. 13, reference numeral 1301 denotes an initial update area limitation for setting a range for selecting an arbitrary solution vector from around each solution vector p _k (k = 1, ..., n) in the solution vector set P. It is a department. The initial solution vector group extracting unit 1302 arbitrarily extracts a solution vector included in the range obtained by the initial update area limiting unit 1301 to obtain a solution vector group P ^k (k = 1, ...,).
n) is created. The solution vector set integration unit 1303 integrates each solution vector group ^Pk into one set P ^* , and the recombination operation is performed on this P ^* .

The operation of the optimization adjusting method and the optimization adjusting apparatus according to the second embodiment of the present invention configured as above will be described with reference to the flowchart of FIG. As a specific problem to be dealt with, the maximum value estimation problem of the multidimensional function w is taken up as in the reference example of the present invention and the first embodiment, and each element constituting the solution vector is also described in FIG. A bit string code is represented by the method, and as the solution vector, the solution vector is represented in a form in which the bit string is arranged according to the order of the elements of the coordinate vector in the multidimensional space.

First, in the initial solution set setting unit 101, an initial set of solution vectors P = [p _k ] (k = 1, ..., n) is set by a predetermined procedure or an instruction from the outside. Limited area radius c # len used in update area limiting section 1301
Is set. Although there are various methods for setting the limited area, the initial update area limiting unit 1301 uses this c # len to set each solution vector p _k (k = 1, ..., In the initial solution vector set P).
Set the spherical region in multidimensional space centered on n). Next, the initial solution vector group extraction unit 1302 causes each solution vector p _k
Set a solution vector group P ^k centered on. In this case, a solution vector group is set by arbitrarily extracting n ^k vectors from each spherical region. Evaluation value acquisition unit 104, fitness calculation unit 1
05 through, each solution in the solution vector set P ^k vectors p ^k
_The fitness f ^k _j (j = 1, ..., n ^k ) is calculated for each _j . The solution vector set integration unit 1303 performs a process of integrating the solution vector groups into one set P ^* based on the goodness of fit. Figure 1
5 schematically shows the situation. As a method therefor, as in the second embodiment, a method of arranging the solution vectors in each solution vector group in descending order of goodness of fit and selecting n ^* from the highest one is conceivable, but in this embodiment, A method of extracting n ^* solution vectors by using the roulette selection method used in the candidate selection unit 201 in the recombination operation unit 108 is adopted. That is, as the probability r ^k _j of the selection of each solution vector p ^k _j , the value r ^k _{j of the} ratio of each fitness f ^k _{j to} the sum f # t is defined as shown in (Equation 10), and the value is defined by the roulette wheel. Assigned to the range occupied by. Then, the method of extracting the solution vector occupying the position where the value selected by the random number stops is used. By doing so, the one with a high degree of suitability is selected stochastically.

[0077]

[Equation 10]

In the recombination operation unit 108, as in the case of the reference example of the present invention, the solution vector p ^* _i (i = 1, ..., In the solution vector set P ^* obtained by the solution vector set integration unit 1303. n ^* )
The selection and selection of the solution vector by the roulette selection method, the crossover process, and the mutation process are performed on the candidate selection unit 201,
This is executed by the crossover process executing unit 202 and the mutation process executing unit 203. This method realizes a local solution update capability by a procedure as simple as possible, as compared with the optimization adjustment apparatus according to the first embodiment of the present invention.

Finally, whether the number of update repetitions l exceeds the number of update end repetitions g # num ^end , or the maximum value f ^max of the goodness of fit f obtained by the goodness-of-fit calculation unit 105 becomes the convergence determination goodness-of-fit f ^end . The end determination is performed based on whether or not the value exceeds the limit. If either is satisfied, the update process ends, and the optimum solution output unit 103 outputs the solution vector with the highest degree of conformity. If both conditions are not satisfied, the process returns to the initial update area limiting unit 1301. The optimal solution estimation is performed by repeatedly executing the above-described processing steps until one of the termination conditions is satisfied. In this way, the third optimization adjusting apparatus of the present invention compares each solution vector in the solution vector set with a vector group randomly extracted from its neighboring space, and newly updates the vector group with a high degree of conformity. Select as a solution vector that belongs to the target solution vector set. Then, by making the resulting solution vector set the target of the recombination operation, it is possible to eliminate the weakness of the local update capability, which was lacking in the conventional genetic algorithm.

An optimization adjusting method and an optimization adjusting apparatus according to the third embodiment of the present invention will be described below with reference to the drawings. FIG. 16 shows the configuration of the optimization adjusting apparatus according to the third embodiment of the present invention. The present embodiment resets the entire solution set by extracting the neighborhood vector group of the solution vector selected based on the goodness of fit, and performs the recombination operation targeting only the solution vector in each neighborhood vector group. The optimum solution is searched for by performing.

In FIG. 16, reference numeral 1601 is a solution set resetting section for resetting the entire solution vector set, 1602 is rearrangement targeting only the solution vector in each neighborhood vector group obtained by the solution set resetting section 1601. It is a group recombination operation unit that generates a new solution vector set by an operation. Here, the solution set resetting unit 1601 selects a representative solution vector from the entire solution vector set, a representative solution vector selecting unit 1603 and a representative solution vector selecting unit 16.
A neighborhood vector group extraction unit 1604 that extracts a neighborhood vector group for each representative solution vector obtained in step 03. The group recombination operation unit 1602 includes a candidate selection unit 201, a crossover processing execution unit 202, and a mutation processing execution unit 203. The candidate selection unit 201 includes a selection range derivation unit 204, a random number generation unit 205, and a solution vector extraction unit 206. Composed.

The operation of the optimizing adjustment method and the optimizing adjusting apparatus according to the third embodiment of the present invention configured as described above will be described with reference to the flowchart of FIG. Here, the specific problem to be dealt with and the method of expressing each element constituting the solution vector are the same as in the case of the reference examples of the present invention and the first and second embodiments.

First, in the initial solution set setting unit 101, an initial set of solution vectors P = [p _k ] (k = 1, ..., n) is set by a predetermined procedure or an instruction from the outside. , The number N _c of representative solution vectors and the number N _{s of} neighborhood vectors extracted centering on each representative solution vector and the neighborhood width w _i (i = 1, ..., N _c ) representing the definition of the neighborhood are set. . The evaluation value acquisition unit 104 obtains the evaluation value E _k of each solution vector according to (Equation 4), and the fitness calculation unit 105 calculates the fitness f _k from the evaluation value E _k .

Next, the operation of the solution set resetting section 1601 will be described. First, the representative solution vector selection unit 1603 selects a representative solution vector p _i ^c (i = 1, ..., N _c ) required to extract a neighborhood vector group from the solution vector set. At this time, various methods can be considered as the selection method, but in the present embodiment, the group recombination operation unit 1602 is used.
A vector with a high probability of suitability is selected as a representative solution vector by the roulette selection method used in the candidate selection unit 201 that configures. The neighborhood vector group extraction unit 1604 extracts N _s neighborhood vectors p satisfying (Equation 11) from the representative solution vector p _i ^c obtained by the representative solution vector selection unit 1603, thereby obtaining N _c Create a set of neighborhood vectors. This (Equation 11) is in the actual real number space.

[0085]

[Equation 11]

Then, the radius w centered on the representative solution vector p _i ^c
_It represents the inside of the sphere of _i , and as a method of selecting N _s neighborhood vectors, it is randomly selected from the solution vectors included in (Equation 11) using uniform random numbers. FIG. 18 is a schematic representation of the above situation. In the present embodiment, the Euclidean distance between the solution vectors in the actual real number space is used as the determination of the neighborhood as in (Equation 11). However, the present invention is not limited to this, and in the two solution vectors x and y converted and expressed in the bit string code, the lower Lbit bits are compared from the right, and the neighborhood is determined based on the number of inverted bits. A method of defining the neighborhood by defining is also considered.

By the processing of the solution set resetting section 1601 as described above, N _c neighborhood vector groups are extracted. The group recombination operation unit 1602 performs a recombination operation targeting only the solution vector in each of the neighboring vector groups. At that time, first, in the candidate selection unit 201, selection of solution vectors by the roulette selection method is performed within each neighborhood vector group, and the crossover process is performed on the new neighborhood vector group obtained by the candidate selection unit 201. Is executed by the crossover processing execution unit 202. In this way, a new solution vector set is generated by the recombination operation independently performed for each neighborhood vector group, and the mutation process execution unit 203 executes the mutation process on this solution vector set.

Finally, whether or not the number of update repetitions l exceeds the number of update end repetitions g # num ^end , or the maximum value f ^max of the goodness of fit f obtained by the goodness-of-fit calculation unit 105 becomes the convergence determination goodness-of-fit f ^end . The end determination is performed based on whether or not the value exceeds the limit. If either is satisfied, the update process ends, and the optimum solution output unit 103 outputs the solution vector with the highest degree of conformity. In this way, a local search can be performed more efficiently by selecting and grouping vector groups from the neighborhood of the representative solution vector. Further, by performing the recombination operation independently for each neighboring vector group, it is possible to avoid that the influence of the solution vector having a high degree of conformity has a large influence on the entire solution immediately.

An optimization adjusting method and an optimization adjusting apparatus according to the fourth embodiment of the present invention will be described below with reference to the drawings. In the fourth embodiment, the neighborhood of the solution vector in the solution vector set is divided into a plurality of regions, and in parallel with the process of arbitrarily selecting from the regions having a high average goodness of fit, in the original solution vector set It is devised to update the solution vector by recombining the solution vector. FIG. 19 shows the configuration of the optimization adjusting apparatus according to the fourth embodiment of the present invention. Update area dividing unit 1
901 performs a process of dividing the neighborhood space of two solution vectors arbitrarily selected from the solution vector set into a plurality of regions, and the average goodness-of-fit calculation unit 1902 makes the update region division unit 190
A plurality of solution vector points are extracted in each area divided by 1, and the evaluation value and the average of the goodness of fit are obtained. Further, the appropriate area solution vector group extraction unit 1903 arbitrarily selects a plurality of solution vectors from the area selected in 1902, and the solution vector integration unit 1904 uses the solution set generated by the recombination operation and the neighborhood area search. The updated solution vector group is integrated into one set P ^* .

The operation of the optimization adjusting method and the optimization adjusting apparatus according to the fourth embodiment of the present invention configured as described above will be described with reference to the flowchart of the process of FIG. The specific problem to be handled and the method of expressing each element forming the solution vector are the same as those in the reference examples of the present invention and the first to third embodiments.

First, as in the above-described embodiments, the initial solution set setting unit 101 sets the initial set of solution vectors P =
[p _k ] (k = 1, ..., n) is set by a predetermined procedure or an instruction from the outside. The evaluation value acquisition unit 104 obtains the evaluation value E _k of each solution vector according to (Equation 4), and the fitness calculation unit 105
But to calculate the goodness of fit f _k than the evaluation value E _k. Recombination operation unit 1
08 performs the genetic recombination operation of the solution vector p _k in the solution vector set P based on this fitness f _k . The following processing is performed in parallel with the processing.

First, in the update area dividing unit 1901, a straight line section L connecting the solution vectors p _k and p _m can be set. This L
Can be expressed as (Equation 12). Where α is a real number.

[0093]

[Equation 12]

This straight line section is divided into the following three areas.
It (1) α <0.0_pmStarting from p_kArea away from (2) Vector p with 0.0 ≦ α ≦ 1.0_kAnd p_mArea between (3) α> 1.0 p_kStarting from p_mArea away from Solution vector p_kFor a random set of solution vectors P
And its solution vector p_mIs selected. like this
Although it is a straight line section, the solution vector p_kSplit the neighborhood of
can do. Note that p_kConsider the sphere area centered on
By dividing the area by the longitude angle β,
A method of dividing into several areas is also conceivable. Average goodness-of-fit calculation unit 1
In 902, n is obtained from each of the areas (1) (2) (3).^cSolution vectors
To calculate the evaluation value and the goodness of fit. And
Average goodness of fit f for each regionⁱ _{me an}to derive (i = 1,2,3)
Is. The appropriate region solution vector extraction unit 1903
Based on the results, select the area with the highest average goodness of fit and
Randomly from here on^dVector based on solution vectors
Group P^kTo extract. Figure 21 shows the situation.
In this case, area (2) is selected. Latest solution vector collection
The solution vector with the highest degree of fit is output. This
The solution vector extracted by the appropriate region solution vector extraction unit 1903 of
Torr Group P^k(k = 1, ..., n) and obtained by the recombination operation unit 108
New solution vector set P is 1904 solution vector integration
One set P in part^*Integrated into. As for this method, the book
Method of extracting in order of high suitability in the second embodiment
And the roulette selection method for the goodness of fit in the third embodiment.
Method to select the one with a high probability of suitability by applying
Can be considered.

As in the case of the above-described embodiments, << termination condition 1 >> Whether the update repeat count l exceeds the update end repeat count g # num ^end << ^end condition 2 >> maximum fitness f _max is determined to converge Two end conditions, that is, whether or not the fitness f ^end is exceeded, are set. If neither is satisfied, the process returns to the evaluation value acquisition unit 104 again, and if either is satisfied, the optimum solution output unit 103 The solution vector with the highest degree of matching is output as the optimum solution. The optimum solution is searched by repeatedly executing the above-described processing steps until the end condition is satisfied. Thus, in parallel with the optimization of the solution vector by the recombination operation in the solution vector set,
Since the update process of the solution vector near each solution vector is also performed at the same time, the local solution update capability can be supplemented to the efficient global solution update capability originally possessed by the genetic algorithm, and the optimal solution vector can be estimated efficiently. It can run well.

The optimization adjusting method and the optimization adjusting apparatus according to the fifth embodiment of the present invention will be described below. FIG. 22 shows the configuration of the optimization adjusting apparatus according to the fifth embodiment of the present invention. In this embodiment, the solution vector set is divided into a plurality of groups based on the arithmetic mean and standard deviation of the goodness of fit, and the optimal solution is searched by performing a recombination operation only on the solution vectors in each group. To execute.

In FIG. 22, reference numeral 2201 denotes a solution set dividing unit which divides the entire solution vector set into a plurality of groups. The solution set division unit 2201 includes a division area determination unit 2202 that determines each area when dividing the entire solution vector set, and a division execution unit 2203 that actually divides the solution vector set into a plurality of groups based on the result. To be done.

The operation of the optimizing adjustment method and the optimizing adjusting apparatus according to the fifth embodiment of the present invention configured as above will be described with reference to the flowchart of FIG. The specific problem to be dealt with and the method of expressing each element forming the solution vector are the same as those in the above-described reference examples and the first to fourth embodiments.

First, the initial solution set setting unit 101
Initial set of vectors P = [p_k] (k = 1, ..., n) is the prescribed procedure
Or it is set by external instructions and the solution
Number of groups to divide as a cuttle set division condition N_gAnd its
The natural constant n used for division is set. Evaluation value acquisition
The evaluation value E of each solution vector is calculated by the unit 104 according to (Equation 4)._kSeeking
Therefore, the goodness-of-fit calculation unit 105 uses the evaluation value E_kMore goodness of fit f_kCalculate
To do. As the convergence condition, the number of update iterations l ends update
Repeat count g # num^endOver, or
Maximum value f of the goodness of fit f obtained by the goodness of fit calculation unit 105^maxBut
Convergence judgment fitness f ^endWhether it exceeds the two
If conditions are set and either is satisfied, the solution
The vector update process is completed, and the optimal solution output unit 103
Also outputs a solution vector with a high degree of conformity.

Next, the operation of the solution set dividing unit 2201 will be described. First, the division width determination unit 2202 calculates the arithmetic mean Ave of the goodness of fit f _k obtained by the goodness of fit calculation unit 105 and its standard deviation σ. Then, for example, in the case of N _g = 3, FIG.
As shown in the figure showing the population distribution for the fitness fk in (a), using Ave, σ and the natural constant n set by the initial solution set setting unit 101, (A) f _k > (Ave + σ × The three areas of the area (B) (A)-(Ave-σ × n) ≦ f _k ≦ (Ave + σ × n) (C) f _k <(Ave-σ × n) are determined. The division executing unit 2203 receives the result of the division area determining unit 2202 and actually divides the entire solution vector set. FIG. 24B is a diagram schematically showing the distribution of each divided group in the actual real value space. Here, in the case of the fourth embodiment of the present invention, the neighborhood vector group of the representative solution vector is extracted regardless of the distribution status examined based on the goodness of fit of the current solution vector set. The embodiment differs greatly in that the current solution vector set is extracted and divided according to the distribution of the goodness of fit. Although n is assumed to be constant in the present embodiment, this value may be dynamically changed.

As in the case of the fourth embodiment of the present invention, the group recombination operation section 1602 uses the solution set dividing section 2
Only the solution vectors in the plurality of groups obtained in 201 are selected by the roulette selection method, and the crossover process is performed by the candidate selection unit 201 and the crossover process execution unit 2
No. 02 is executed in order. Then, the mutation process executing unit 203 executes the mutation process on the entire newly generated solution vector set.

The above-described update process of the solution vector set is repeatedly performed until the above-mentioned convergence condition is satisfied, thereby estimating the toll for an optimum solution. By dividing into multiple groups based on the mean and standard deviation of each, and performing the recombination operation targeting only the solution vectors belonging to each group independently, The low solution vector is divided into different groups and the recombination operation is performed. Therefore, it is possible to avoid that they have a great influence on the entire solution vector set immediately, and also to speed up the convergence of the solution in each group, so that the optimal solution vector can be estimated efficiently.

An optimization adjusting method and an optimization adjusting apparatus according to the sixth embodiment of the present invention will be described below with reference to the drawings. In the sixth embodiment, every time the goodness of fit is calculated, it is determined whether or not the solution vector set is divided into a plurality of groups based on the arithmetic mean and the standard deviation of the goodness of fit and the maximum and minimum goodness of fit. An optimal solution search is executed by selecting a recombination operation target by and performing a recombination operation targeting the solution vector in the obtained recombination operation target. FIG. 25 shows the configuration of the optimization adjusting apparatus according to the sixth embodiment of the present invention, in which 2501 represents the entire solution vector set based on the arithmetic mean and standard deviation of the goodness of fit and the maximum and minimum goodness of fit. It is a rearrangement target control unit that determines whether to divide into a plurality of groups.

The operation of the optimization adjusting method and the optimization adjusting apparatus according to the sixth embodiment of the present invention configured as described above is a flowchart showing the overall processing of FIG. 26, and processing 2 of FIG. Description will be given based on a flowchart. The specific problem to be handled and the method of expressing each element forming the solution vector are the same as those in the reference examples of the present invention and the first to fifth embodiments.

First, as in the above-described embodiments, the initial solution set setting unit 101 sets the initial set of solution vectors P =
[p _k ] (k = 1, ..., n) is set by a predetermined procedure or an instruction from the outside, and the number of groups N _g to be divided as a solution vector set division condition and a natural constant used for the division n is set. The evaluation value acquisition unit 104 obtains the evaluation value E _k of each solution vector according to (Equation 4), and the fitness calculation unit 10
5 calculates the fitness fk from the evaluation value E _k . As the convergence condition, the update repeat count l is the update end repeat count g # num
Whether or not it exceeds the ^end , or the fitness calculation unit 10
The maximum value f ^max of the fitness f obtained in 5 is the convergence determination fitness f
Two conditions are set to determine whether or not ^end is exceeded. If either of them is satisfied, the solution vector update process ends, and the optimal solution output unit 103 outputs the solution vector with the highest degree of conformity.

In this embodiment, a control mechanism 2501 for determining whether or not to divide the solution vector set is provided in the fifth embodiment, so that meaningless division (fitness of all solution vectors in the solution vector set) is achieved. The purpose is to avoid the loss of efficiency due to The recombination target control unit 2501 first obtains the average Ave and standard deviation σ thereof from the goodness of fit f _k obtained by the goodness of fit calculation unit 104. Then, based on this value and the maximum / minimum goodness of fit, it is determined whether or not the solution vector set is divided into a plurality of groups. Various criteria can be considered as the criterion, but in the present embodiment, it is defined that it is necessary to divide the solution vector set into a plurality of groups when the condition expressed by (Equation 13) is satisfied. To do.

[0107]

[Equation 13]

Here, f _max is the maximum fitness, f _min is the minimum fitness, and c _f is the fitness distribution judgment constant. The 2501 performs the following operation according to the result of this determination criterion. << Determination 1 >> When it is determined that the division is not performed, the entire solution vector set is targeted for recombination and the process proceeds to the recombination operation unit 108. << Determination 2 >> If it is determined that the division is performed, the process proceeds to the solution set division unit 2201. At this time, a method similar to the case of the sixth embodiment of the present invention is used as a method of dividing the solution vector set based on the average value Ave of the goodness of fit and the standard deviation σ.

In the recombination operation unit 108, the selection selection of the solution vector by the roulette selection method for the recombination target obtained by the recombination target control unit 2501 is performed by the candidate selection unit 2
01, the one-point or two-point crossover process is executed by the crossover process execution unit 202, and the mutation process is executed by the mutation process execution unit 203.

Then, it is determined whether or not to continue updating the solution vector set under the same convergence condition as in the reference example of the present invention and the first to fifth embodiments. The optimal solution vector is estimated by repeatedly executing the above processing steps. In this way, it is determined whether or not the solution vector set is divided into a plurality of groups based on the arithmetic mean and standard deviation of the goodness of fit and the maximum and minimum goodness of fit. If it is determined to be divided, the solution set division unit divides the entire solution set into a plurality of groups, and if not so, the entire solution set is the target of the recombination operation, and the solution for the obtained recombination processing target is solved. By selecting a vector and performing a recombination operation, it is determined whether there is a solution vector with a high degree of goodness of conformity based on the goodness of fit distribution situation, and by limiting the recombination processing target so as to reduce the effect, efficiency can be improved. A good optimal solution vector estimation can be performed.

The optimization adjusting method and the optimization adjusting apparatus according to the seventh embodiment of the present invention will be described below. FIG. 28 shows the configuration of the optimization adjusting apparatus according to the seventh embodiment of the present invention. In the present embodiment, the step convergence conformity and the number of solution vector individuals that satisfy the value are defined as the step convergence criterion used in the update area setting unit, and updated every time the update area setting unit satisfies the step convergence criterion. The optimal solution vector is estimated by locally limiting the region and increasing the accuracy of the stepwise convergence fitness.

In FIG. 28, reference numeral 2801 evaluates the graded convergence evaluation value set as the graded convergence criterion and the number of solution vector individuals satisfying the criterion, and when the criterion is satisfied, the search area is locally limited and The update area setting unit 2 is an update area setting for increasing the accuracy of the stepwise convergence adaptability.
Reference _numeral 801 denotes a step convergence determination unit 2802 that evaluates the step convergence suitability fl _th and the number of solution vector individuals Nl _th that satisfies the value.
And a convergence criterion changing unit 2803 that locally limits the update region and improves the accuracy of the gradual convergence suitability when the gradual convergence determination unit 2802 determines convergence.

The operation of the optimization adjusting method and the optimization adjusting apparatus according to the seventh embodiment of the present invention configured as above will be described with reference to the flowchart of FIG. At this time, as the specific problem to be dealt with and the coding method of the solution vector, it is assumed that the same as those in the embodiments of the reference example and the first to sixth embodiments are used.

First, as in the previous embodiments, the initial solution set setting unit 101 sets the initial set of solution vectors P = [p
_k ] (k = 1, ..., n) is set by a predetermined procedure or an instruction from the outside. Optimal solution vector estimation is started from this initial solution vector set P. At the same time, the update region of the initial solution vector and the initial stage convergence criterion are set. Phase convergence criteria are defined as the solution vector population Nl _th which satisfies the phase convergence adaptability fl _th and the value, the initial population N for convergence determined initial stage converging fit fl _th0
l _th0 is set as fl _th = fl _th0 , Nl _th = Nl _th0 . Also, as the initial update area, only the most significant Clen bit is effective in the binary code of length Blen corresponding to the coefficient, and if it is less than that, the area expressed by being regarded as 0 is targeted.

In the update area setting section 2801, first, in the step convergence determination section 2802, the goodness of fit f _k obtained at 105 is calculated.
And the stepwise convergence fitness fl _th are compared, and the number Nl of solution vectors for which f _k is larger than fl _th is counted. At this time, the staged convergence determination unit 2802 sets a staged convergence condition for determining whether or not the staged convergence fitness fl _th exceeds a preset convergence determination fitness f ^end, and makes the following determination. << Judgment 1 >> When Nl ≧ Nl _th and the step convergence condition are satisfied,
Proceed to the optimum solution output unit 103. << Judgment 2 >> When Nl ≧ Nl _th but the stepwise convergence condition is not satisfied, the convergence criterion changing unit 2803 changes the stepwise convergence adaptability fl _{th according} to (Equation 14). Here, fl _th ^new is the newly set stepwise convergence fitness, fl _th ^prv is the current stepwise convergence fitness, and Δfl is the variation of the preset stepwise convergence fitness, and a positive real constant. Is. Here, △ fl
Is considered to be constant, but it is possible to set this value so that it also changes dynamically.

[0116]

[Equation 14]

In the convergence criterion changing unit 2803, the Clen bit, which is the current update area of each binary-converted coefficient, is
If the coefficient is included in the vector satisfying fl _th ^prv , it is saved as it is. On the other hand, for the dissatisfied vector, a vector randomly selected from the satisfied vectors is Clen which is the current search area of each coefficient.
Copied to bit. After that, the area represented by the next lower Clen bit in each coefficient is again set as the update target area. At that time, the bits above the update target area are not changed, and the bits below the update area are set to 0. Here, when the movement of the Clen bit exceeds the preset Blen bit assigned to each coefficient, the Clen bit is updated from the least significant bit. After the above processing, the determination by the step convergence determination unit 2802 is repeated again. << Judgment 3 >> If Nl ≦ Nl _th , the process proceeds to the recombination operation unit 108.

FIG. 30 shows the update bit selection process performed in the determination 2 expanded into an update area in the real number space that can be taken by the actual coefficient. Since the coefficient that is an element of the solution vector is expressed in the bit string code as shown in FIG. 7, the fluctuation of the upper bit corresponds to the large fluctuation of the coefficient in the actual real number space, and the fluctuation of the lower bit is centered on the current point. Corresponds to the local fluctuation. Therefore, as is clear from FIG. 30, the update bit selection process performed in the determination 2 is a process of first globally updating the region that each coefficient can take, and then further subdividing the extracted region and locally updating it. This is equivalent to the process of gradually updating and limiting the update area, such as repeating.

As the convergence condition for judging the end of the update of the solution vector in the solution vector set, the step convergence condition used in judgment 1 and judgment 2 and the update repetition number l exceed the update end repetition number g # num ^end . There is a condition to judge whether or not. The optimum solution is estimated by repeatedly executing the above-described solution vector update processing until these two conditions are satisfied. In this way, by first performing a search for a globally rough solution vector and then gradually shifting to a local search, it is possible to reduce the effect of the poor local search ability in the neighborhood of the optimal solution, which has been a problem in the past. It is possible to realize an efficient and high-speed optimal solution search.

Reference examples in the present invention, first to seventh
In the optimization adjusting method and the optimization adjusting apparatus by the genetic algorithm of the embodiment of the present invention, the maximum value estimation problem of the multidimensional function has been described as a specific example, but the traveling salesman problem and the capacity of a certain number of items are limited. Knapsack problem that seeks to pack as many knapsacks as possible, multidimensional function approximation problem that finds a multidimensional function that approximates them well from multiple points in a given multidimensional space, nurse night shift in hospital, midnight work schedule It is also possible to apply it to the scheduling problem to properly plan the aircraft, parametric design of aircraft, etc. 1
An example is the one-dimensional function approximation problem that can be described as follows. << Condition 1 '>> A set Q composed of t two-dimensional vectors q _i (x _i , y _i ) (i = 1, ..., t) is created by a predetermined procedure or an instruction from the outside. Define a (s-1) -order function such as << Condition 2 '>> (Equation 15). << Condition 3 '>> The coefficient a _j (j = 0, ..., s-) in each term of the (s-1) -dimensional function represented by (Equation 15) that most approximates the two-dimensional vector set Q in Condition 1 1) is calculated within the range that satisfies (Equation 16). Where x ∋ x _k is the independent variable and c _j is the coefficient a _j
Considered as the upper search limit for the absolute value of.

[0121]

[Equation 15]

[0122]

[Equation 16]

_< Condition 4> However, it is assumed that the number t of two-dimensional vector data in Condition 1 is much larger than the number s of coefficients a _{j to be} searched. FIG. 31 is a schematic representation of this. In this case, the evaluation value acquisition unit 104 may give the evaluation value in advance, for example, as in (Equation 17).

[0124]

[Equation 17]

Vector data o _i (x _i , z) estimated from the obtained two-dimensional vector data q _i (x _i , y _i ).
A method of obtaining an evaluation value E _k for each solution vector p _k by using the sum of squared errors (i = 1, ..., t) between _i ) can be considered. The problem can be regarded as a minimization problem that minimizes this evaluation value. At that time, the fitness calculation unit 105 calculates the fitness from the evaluation value obtained by the evaluation value acquisition unit 104 in order to see the suitability of each solution vector. Various functions can be considered as a function for deriving the fitness f _k , but here, f _k = VE
and _k, V is the E _k greater than the positive real constant. By doing so, the smaller the evaluation value, the larger the adaptability can be converted.

Hereinafter, first to first examples related to the present invention will be described.
The optimization adjustment method and the optimization adjustment device in the fifth example will be described with reference to the drawings. The present invention relates to an optimization adjusting method and an optimization adjusting device using a genetic algorithm.

The optimization adjusting method and the optimization adjusting apparatus in the first example related to the present invention will be described with reference to the drawings. The optimization adjustment method and the optimization adjustment apparatus of the first example limit the update area of each solution vector to be adjusted based on its characteristics and past adjustment results, and the interactive genetic method is applied to the solution vector in the area. The optimal solution vector is adjusted by the genetic algorithm. Example 1 is an example in which the optimization adjustment method and the optimization adjustment device are applied to the adjustment of a lens for correcting visual acuity.

In FIG. 32, 3201 is a target data input unit for inputting target data, 3202 is an update region limiting unit for limiting the region that a solution vector can take, 101 is a set of initial solution vectors to be adjusted P = [ An initial solution set setting unit 3203 for setting p _k ] (k = 1, ..., n) by a predetermined procedure or an instruction from the outside is a main that actually adjusts the optimum solution vector by genetic recombination operation or the like. Processing unit, 10
3 is an optimum solution output unit for outputting an optimum solution vector, 320
Reference numeral 4 denotes a device adjustment execution unit that adjusts the device that handles the data input by the target data input unit 3201 according to the optimum solution vector output by the optimum solution output unit. The main processing unit 3203 presents the information represented by each solution vector to the user and asks the user to evaluate each solution vector 3205.
And a goodness-of-fit calculation unit 105 that calculates a goodness-of-fit f _k for each problem of each solution vector based on the evaluation value E _k obtained by the user evaluation unit 3205, and each solution obtained by the goodness-of-fit calculation unit 105. A recombination operation unit 108 that executes a recombination operation process such as selection and crossover and mutation of a solution vector based on the goodness of fit of the vector, and an update area limiting unit in the solution vector set obtained by the recombination operation unit 1081. A set resetting unit 32 that resets a set of solution vectors by replacing a solution vector set in 3202 that does not belong to a limited region that can be taken by an arbitrary vector in the limited region
06. The user evaluation unit 3205 provides the information presentation unit 32 that presents the information represented by each solution vector to the user.
07 and the user evaluation determination unit 3208 that determines the evaluation value of each solution vector based on the information presented by the information presentation unit 3207 by the user. Recombination operation unit 108
2, is composed of a candidate selection unit 201, a crossover processing execution unit 202, and a mutation processing execution unit 203. The candidate selection unit 201 includes a selection range derivation unit 204, a random number generation unit 205, and a solution vector extraction. It is configured by the unit 206.

The operation of the optimization adjusting method and the optimization adjusting apparatus in the first example related to the present invention configured as above will be described with reference to the flowchart of FIG.
First, as shown in FIG. 72, the target data input unit 32
At 01, the test image to be presented to the user is input as input data. The solution vector p _{k to be} adjusted is (the power L ^SPH (D) of the spherical lens used in the left eye, the power L ^CYL (D) of the cylindrical lens for astigmatism, the angle L ^AXIS (degree) of the axis of astigmatism, and used in the right eye) It is represented by a set of spherical lens power R ^SPH (D), astigmatism cylindrical lens power R ^CYL (D), and astigmatic axis angle R ^AXIS (degrees). Here, the lens power is the reciprocal of the focal length (m).
It has D as a unit. Then, myopia is represented by a sign of (-) and hyperopia (+). For example, in the case of myopia where the far point of the focal length is 50 cm, -2.0D (1 m ÷ 50 cm) is displayed, and the power of the lens to be corrected is also- 2.0D is used. It should be noted that although it is assumed that the solution vector has a real value as an element, FIG.
It is also conceivable to convert each real value into a corresponding bit string code of length Blen and express the bit string code in order. As shown in the conceptual diagram of FIG. 73, the update region limiting unit 3202 limits the search to, for example, only the region Φ occupied by the past optimum solution vector in the solution vector space. Initial solution set setting unit 101
Then, the solution vector in this region Φ is randomly selected using uniform random numbers, and the initial set of solution vectors P = [p _k ] (k =
1, ..., n) is set. Then, the adjustment of the optimum solution vector is started from this initial solution vector set P. The information presenting unit 3207 presents an adjustment image viewed through the test image input by the target data input unit 3201 to the correction lens represented by the solution vector p _k that is the source of the above-described solution vector set P, and sees this. Then, the user performs the evaluation value E _k of the correction lens created by each solution vector in the user evaluation determination unit 3208. The goodness-of-fit calculation unit 105 calculates the goodness-of-fit f _k for checking the suitability of each solution vector from the evaluation value obtained by the user evaluation determination unit 3208. Goodness of fit
Various functions have been proposed as a function for deriving f _k , but here, f _k = E _k .

The operation of the recombination operation unit 108 is the same as that of the first to eighth embodiments of the present invention, and therefore its explanation is omitted. The set resetting unit 3206 takes out a solution vector that does not belong to the limited region Φ set by the update region limiting unit 3202 from the solution vectors newly obtained by the recombination operation, and randomly selects it from the limited region Φ. The solution vector is reset by adding the generated solution vector. As a termination condition, it is determined whether the number of repetitions does not exceed the allowable number of repetitions, and it is also confirmed whether the information requested by the user for evaluation is satisfactory to the user. If neither is satisfied, the process returns to the information presenting unit 3207 again. Optimization of the solution vector is realized by repeatedly executing the above-described processing steps until the end condition of the number of repetitions or the satisfaction of the user is satisfied. Then, the optimum solution output unit 103 outputs the solution vector having the highest evaluation value as the optimum solution vector, and the device adjustment execution unit 3204 adjusts or instructs the correction lens according to the optimum solution vector. By performing such processing, it becomes possible for the user to easily adjust the lens for correcting the visual acuity according to the appearance based on the visual acuity of the user. Then, the search area of the solution vector to be adjusted is limited based on its characteristics and past adjustment results, and the solution vector in that area is targeted, so that the search in the area that does not need to be searched is deleted. Therefore, the optimum solution vector can be quickly adjusted.

The optimization adjusting method and the optimization adjusting apparatus in the second example related to the present invention will be described below with reference to the drawings. In the second example, the optimization adjusting method and the optimization adjusting apparatus set the initial set of solution vectors to be adjusted based on the recorded past optimum solution vector information, and use the interactive genetic algorithm to optimize the set. The solution vector is derived. The second example is the first related to the present invention.
This is an example applied to adjustment of a lens for correcting visual acuity as in the case of. FIG. 34 shows the configuration of the adjusting device in the second example related to the present invention.

In FIG. 34, reference numeral 3401 denotes a recording medium unit in which the optimum solution vector obtained by past adjustment is recorded, and 3402 reads the stored optimum solution vector group from the recording medium unit 3401. A recorded information reading unit 3403 is an initial solution vector selection unit 3404 that selects one used for the initial set P of solution vectors from the past optimum solution vectors read by the recorded information reading unit 3402.
Is an initial solution vector replenishment unit that adds an randomly set solution vector to the solution vector selected by the initial solution vector selection unit 3403 to set an initial set of solution vectors, 3405.
Is an optimal solution vector recording unit that records the optimal solution vector output by the optimal solution output unit 103 for the user in the recording medium unit 3401.

The operation of the optimization adjusting method and the optimization adjusting apparatus in the second example related to the present invention configured as above will be described with reference to the flowchart of FIG.

First, in the target data input section 3201, the test image data to be shown to the user is input. The solution vector p _{k to be} adjusted is the same as in the case of the optimization adjusting method and the optimization adjusting device in the first example related to the present invention (the power L ^SPH (D) of the spherical lens used for the left eye, Cylindrical lens power L ^CYL (D), astigmatic axis angle L ^AXIS (degrees), spherical lens power R ^SPH (D) used for the right eye, cylindrical lens power R ^CYL (D), astigmatism It is represented by the set of axis angles R ^AXIS (degrees). The recording information reading unit 3402 is a recording medium unit 340.
The optimum solution vector group obtained by the past adjustment recorded in 1 is read. Initial solution vector selection unit 3403
But, n _s number selected from historical optimum solution vector group obtained by the recorded information read unit 3402, following the initial solution vector refill unit 3404 (nn _s) number of solutions vector set randomly by uniform random numbers To create an initial set P of solution vectors. If the recording medium section 3401 does not have n _s pieces of information, only the recorded solution vectors are selected by the initial solution vector selection section 3403, and the original number of the initial set of solution vectors becomes n pieces. Initial solution vector replenishment unit 3
It is set in 404. Presenting an adjusted image through the correction lens in the information presenting unit 3207, determining the evaluation value of each solution vector in the user evaluation determining unit, and the fitness calculating unit 1
In 05, the degree of fitness of each solution vector is calculated, and the recombination operation unit 108 performs the recombination operation of the solution vectors to generate a new solution vector set.

The termination condition is the first condition related to the present invention.
As in the case of the optimization adjustment method and the optimization adjustment device in the example above, determine whether the number of iterations exceeds the allowable number of iterations, and whether the information requested by the user is satisfactory to the user. Check to see if it is possible. If neither is satisfied, the process returns to the information presenting unit 3207 again. Optimization of the solution vector is realized by repeatedly executing the above-described processing steps until the end condition of the number of repetitions or the satisfaction of the user is satisfied. Then, the optimum solution output unit 103 outputs the solution vector having the highest evaluation value as the optimum solution vector, and the device adjustment execution unit 3204 adjusts or instructs the correction lens according to the optimum solution vector. At this time, the optimum solution vector is recorded in the recording medium section 3401 by the optimum solution vector recording unit 3405 so as to be the starting point for the next adjustment.

By performing such processing, the user can easily adjust the lens for correcting the visual acuity according to the appearance based on the visual acuity of the user. Furthermore, the optimum solution vector is adjusted efficiently by setting an initial set of solution vectors based on the recorded optimum adjustment information in the past and excluding and searching for solution vectors that are not preferable to the user. Therefore, it is considered that the burden on the user when evaluating each solution vector is reduced.

The optimization adjusting method and the optimization adjusting apparatus in the third example relating to the present invention will be described below with reference to the drawings. The optimization adjusting method and the optimization adjusting device of the third example correct the evaluation value of the user based on the psychological condition estimated based on the physiological information of the user, and interactively based on the corrected evaluation value. The optimal solution vector is adjusted by the genetic algorithm. Example 2 deals with the adjustment problem of the lens that performs the correction of the visual acuity like the first example related to the present invention. FIG. 36 shows the configuration of the optimization adjusting apparatus in the third example related to the present invention.

In FIG. 36, reference numeral 3601 is a user psychological estimation unit for estimating the psychological situation of the user from physiological data when the user evaluates the information represented by each solution vector, and 3602 is the psychological situation of the user based on the obtained psychological situation. It is an evaluation value correction unit that corrects the determined evaluation values. User psychological estimation unit 36
01 is a physiological data measuring unit 360 that actually measures physiological data when the user evaluates the information represented by each solution vector.
3 and a psychological estimation execution unit 3604 that estimates the psychological state of the user based on the physiological data.

The operation of the optimization adjusting method and the optimization adjusting apparatus in the third example related to the present invention configured as above will be described with reference to the flowchart of FIG.

First, the test image data to be shown to the user is input in the target data input section 3201. The solution vector p _{k to be} adjusted is the same as in the case of the optimization adjusting method and the optimization adjusting device in the first example related to the present invention (the power L ^SPH (D) of the spherical lens used for the left eye, Cylindrical lens power L ^CYL (D), astigmatic axis angle L ^AXIS (degrees), spherical lens power R ^SPH (D) used for the right eye, cylindrical lens power R ^CYL (D), astigmatism It is represented by the set of axis angles R ^AXIS (degrees). In the initial solution set setting unit 101, an initial solution set P = [ _pk ] (k = 1, ..., n) of solution vectors is set by a predetermined procedure or an instruction from the outside. Information presenting section 320
In 7, the adjustment image viewed through the correction lens formed by each solution vector is presented to the user, and the user evaluates the appearance of the adjustment image by each solution vector in the user evaluation determination unit 3208. At the same time, the physiological data measuring unit 3603 measures the physiological data of the user when evaluating the adjusted image with each solution vector. As the physiological data to be measured, various data (skin electrode resistance, respiration rate, pulse, blood pressure, eye movement, brain wave, etc.) can be considered. Here, an example in which the number of eye blinks and the amount of sweating are used will be described.

As shown in FIG. 75 (a), in the case of human beings, blinking is suppressed and the frequency of blinking becomes lower than usual when the subject being viewed is interesting, interested, or attracting attention. . Then, when released from such a state, this time it frequently happens as if to recover the suppressed amount, and eventually it returns to the original. By using this, the user's blink occurrence frequency n _eye is set to a preset allowable blink frequency n _eye.
If it is larger than ^th, it can be determined that the interest of the user is low. Further, as shown in FIG. 6B, when the person is nervous or agitated, the amount of perspiration increases and the skin electrode resistance fluctuates greatly. The psychological estimation execution unit 3604 makes these determinations. First, when the psychological estimation execution unit 3604 determines that the user's interest is low, the solution vector p _k
For the evaluation value E _k (k = 1, ..., n) of, the average evaluation value E _ave and the standard deviation σ _E are obtained, and the following correction is performed. (i) If 0.0 <E _k -E _ave ≤ σ _E , then E _k ← E _k + correction amount △ E _cor ¹ (ii) If 0.0 <E _ave -E _k ≤ σ _E , then E _k ← E _k -Correction amount ΔE _cor ¹ Generally, when the interest is low, the fluctuation of the user evaluation value is likely to be small.
It is possible to distribute _k widely, which is effective for adjusting the optimal solution vector by the genetic algorithm. If it is determined that the subject is upset, (iii) if E _k -E _ave ≥σ _E is satisfied, then E _k ← E _k -correction amount ΔE
_cor ² (iv) E _ave -E _k ≧ σ _E , E _k ← E _k + correction amount ΔE
Correct _cor ² . Contrary to the corrections of (i) and (ii), this is expected to have the effect of suppressing the fact that the fluctuation of the user evaluation tends to become large due to the influence of shaking. From (i) to (i
In the correction of v), when E _k > maximum possible evaluation value EA _max , E _k = E
And A _max, and E _k = EA _min when E _k <minimum possible evaluation value EA _min.

The fitness calculation section 105 includes an evaluation correction section 3602.
The goodness of fit is calculated based on the corrected evaluation value obtained in step S3, and the recombination operation unit 108 performs an operation of recombination of solution vectors for selective selection, crossover processing, and mutation processing. The process from the presentation of the adjusted image to the rearrangement operation in the presentation information section is
It is repeatedly executed until the number of repetitions exceeds the allowable number of repetitions or the information requesting the user to evaluate is created enough to satisfy the user. And the optimal solution output unit 1
The solution vector with the highest evaluation value in 03 is output as the optimum solution vector, and the device adjustment execution unit 3204 adjusts or instructs the lens for vision correction according to the optimum solution vector. In this way, the psychological state of the user is always estimated based on the physiological data when the user is performing the evaluation, and the evaluation value of the user is corrected based on the estimation result. It is possible to reduce the influence, and it is possible to realize the optimum solution vector adjustment for the user's situation (hearing, visual acuity, etc.) without being influenced by the environment.

The optimization adjusting method and the optimization adjusting apparatus in the fourth example related to the present invention will be described below with reference to the drawings. The fourth optimization adjusting method and the optimization adjusting device estimate the evaluation model of the user's adjustment process based on the adjustment history when the user searches for the optimum solution vector using the interactive genetic algorithm. To do. Example 3 deals with the adjustment problem of the lens that performs the correction of the visual acuity similarly to the first example related to the present invention. FIG. 38 shows the configuration of the adjusting device in the fourth example related to the present invention. Figure 3
8, reference numeral 3801 denotes an evaluation model output unit for outputting a set of parameters capable of expressing the evaluation model regarding the user's adjustment process obtained by the main processing unit 3203,
Reference numeral 3802 denotes a second user evaluation unit that performs processing such as presenting the information represented by each solution vector to the user for evaluation by the user and recording the solution vector corresponding to the evaluation value.
3803 is an evaluation model estimation determination unit that determines whether or not the recorded user adjustment history satisfies a preset evaluation model estimation condition, and 3804 is an evaluation model estimation determination unit 3803 that satisfies the evaluation model estimation condition. If so, the model estimation execution unit estimates the evaluation model regarding the adjustment process of the user based on the adjustment history of the user recorded in the adjustment history recording unit. The second user evaluation unit presents information of each solution vector to the user, an information presentation unit 3207, a user evaluation determination unit 3208 that requests the user to evaluate each solution vector from the information presented by the information presentation unit 3207, User evaluation determination unit 3208
The adjustment history recording unit 3805 is configured to record the solution vector corresponding to the evaluation value obtained in (3).

The operation of the adjusting device having the above-mentioned structure according to the fourth example of the present invention will be described with reference to the flowchart of FIG.

First, test image data to be shown to the user is input in the target data input section 3201. The solution vector p _{k to be} adjusted is the same as in the case of the optimization adjusting method and the optimization adjusting device in the first example related to the present invention (the power L ^SPH (D) of the spherical lens used for the left eye, Cylindrical lens power L ^CYL (D), astigmatic axis angle L ^AXIS (degrees), spherical lens power R ^SPH (D) used for the right eye, cylindrical lens power R ^CYL (D), astigmatism It is represented by the set of axis angles R ^AXIS (degrees). In the initial solution set setting unit 101, an initial solution set P = [ _pk ] (k = 1, ..., n) of solution vectors is set by a predetermined procedure or an instruction from the outside. Information presenting section 320
In 7, the adjustment image viewed through the correction lens formed by each solution vector is presented to the user, and the user evaluates the appearance of the adjustment image by each solution vector in the user evaluation determination unit 3208. Adjustment history recording unit 380
5 records the solution vector corresponding to the evaluation value obtained by the user evaluation judgment unit 3208. Then, the evaluation model estimation determination unit 3803 determines whether or not the evaluation model estimation condition is satisfied. Although various evaluation model estimation conditions are possible, the number of repetitions counted as one when the loop is performed through the processing from the second user evaluation unit 3802 to the shuffle operation unit 108, lo
When op ^th is defined as a preset allowable number of iterations, loop ≧ loop ^th is used as an evaluation model estimation condition. this is,
It is based on the assumption that more than a certain amount of data will be needed to estimate the evaluation model for the user's adjustment process.

When the evaluation model estimation determination unit 3803 does not satisfy the evaluation model estimation condition of the loop count, the fitness calculation unit 105 calculates the fitness of each solution vector and the recombination operation unit 108 rearranges the solution vectors. The operation is performed to generate a new solution vector set, and the information presenting unit 3207 is again based on the new solution vector set.
Then, the adjusted image viewed through the correction lens formed by each solution vector is presented to the user. On the contrary, when the evaluation model estimation determination unit 3803 satisfies the evaluation model estimation condition of the number of iterations loop, the model estimation execution unit 380
In 4, the estimation history of this user is estimated using the history of the adjustment of the user recorded in the adjustment history recording unit 3805. Various methods can be considered as the estimation method of the evaluation model related to the adjustment process of the user, but here, a neural network as shown in FIG. 76 is used. The neural network of this example is a feedforward type network composed of an input layer, an intermediate layer, and an output layer as shown in the figure. Input layer is 6 × n
It consists of neurons. Solution vector p _k =
(L ^SPH _k , L ^CYL _k , L ^AXIS _k , R ^SPH _k , R ^CYL _k , R ^AXIS _k ) (k =
(1, ..., n) component values and (n-1) vector differences p _k -p _l =
(L ^SPH _k -L ^SPH _l , L ^CYL _k -L ^CYL _l , L ^AXIS _k -L ^AXIS _l , R ^SPH _k -R
^SPH _l , R ^CYL _k -R ^CYL _l , R ^AXIS _k -R ^AXIS _l ) (k, l = 1, ..., n, l ≠
The value of the component of k) is input. The output layer consists of n neurons corresponding to the ranks in the solution vector set with n elements. The output of the input layer is transmitted to the intermediate layer and the output layer, and the output of the network is obtained. The output o _{i of} each neuron is calculated according to the following (Equation 18).

[0147]

[Equation 18]

In (Equation 18), o _i is an output of each neuron i, w _ij is a connection weight, x _j is an input from another neuron, and θ _i is a threshold. The nonlinear function f is
It is a sigmoid function shown in (Equation 19).

[0149]

[Formula 19]

Each neuron performs the calculation shown in (Equation 18) and (Equation 19) and outputs the result.

The model estimation execution unit 3804 carries out learning of this neural network so that the adjustment process of the user can be absorbed. In other words, the order of p _k in the solution vector set can be determined from the difference component 6 × (n-1) between the 6 components of the solution vector p _k and the other solution vector p _l in the solution vector set. The connection weight of the network is changed (learned) so that it can be done. At that time, by giving 1 as a teacher signal to the output neuron corresponding to the rank of the solution vector obtained from the evaluation value recorded in the adjustment history recording unit 3805 and 0 as a teacher signal to the other output neurons, Learning is performed, and in the direction of reducing the error between the output signal and the teacher signal shown in (Equation 20),
The backpropagation method that changes the connection weight of each neuron as in (Equation 21) is used as a learning method.

[0152]

[Equation 20]

Here, out is the output signal of the network, t
arget is the teacher signal.

[0154]

[Equation 21]

Here, Δw _ij (n) is the degree of change of the coupling load,
α and η are appropriate positive real numbers, n ^loop is the number of learnings, and ∂Err /
∂w _ij represents the sensitivity of the change in each coupling weight to the output error of the network. In (Equation 21), the first term is the load changing direction that reduces the error, and the second term is the inertia term. The value w _ij of each connection weight of the neural network obtained by learning in this way is used as the evaluation model output unit 3
It is output by 801. As described above, the interactive genetic algorithm can be used to estimate the evaluation model for the user's adjustment process based on the adjustment history when the user adjusts the optimal solution vector. The model can also be used to estimate the inverse of the solution vector given. Although the neural network in this example applies the learning method by backpropagation to the neuron using the sigmoid function, the present invention is not limited to this learning method. For example, application of a learning method using a conjugate gradient method, a quasi-Newton method, or the like can be considered. In addition to the three-layer feedforward type network,
It is also possible to apply a network using the learning vector quantization described in the example of 1. and a neural network having a feedback connection from the output layer to the input layer.

The optimization adjusting method and the optimization adjusting apparatus in the fifth example related to the present invention will be described below with reference to the drawings. The fifth optimization adjusting method and the optimization adjusting device related to the present invention are adjusted by a plurality of users based on an adjustment history when a solution vector is optimized by a plurality of users using an interactive genetic algorithm. It estimates a common model of the process. Example 5 deals with the adjustment problem of the lens that performs the correction of the visual acuity similarly to the first example related to the present invention. FIG. 40 shows the configuration of the adjusting device in the fifth example related to the present invention. In FIG. 40, reference numeral 4001 denotes a common model output unit that outputs a set of parameters that can be expressed by the main processing unit 3203 and that can express a common model relating to the adjustment process of a plurality of users, and 4002 satisfies a condition for ending adjustment by a user. A user adjustment end determination unit 4003 that determines whether or not to perform the adjustment, and a common model estimation unit 4003 that estimates a common model regarding the adjustment process from the obtained adjustment history of a plurality of users. Common model estimation unit 4003
Is a common model estimation determination unit 4004 that determines whether or not the adjustment history of a plurality of users recorded in the adjustment history recording unit 3805 satisfies a preset common model estimation condition, and a common model estimation condition. A common model estimation execution unit 4005 that actually estimates the common model regarding the adjustment process of a plurality of users when satisfied.

The operation of the optimization adjusting method and the optimization adjusting apparatus in the fifth example related to the present invention configured as above will be described with reference to the flow charts of FIGS. 41 and 42.

First, the target data input section 3201 is used.
The test image data to be shown to the user is input. Solution to adjust
Cutle p_kIs the optimization adjustment method in the first example of the present invention.
This is similar to the case of the optimization adjusting device. That is, the solution vector
Le is p_k= (L^SPH _k, L^CYL _k, L^{AXI S} _k, R^SPH _k, R^CYL _k,
R^AXIS _k). In the initial solution set setting unit 101
, The initial set of solution vectors P = [p_k] (k = 1, ..., n) is set
Is determined. In the information presenting unit 3207, each solution vector
The adjustment image that can be seen through the correction lens created by
User is presented to the user evaluation determination unit 3208.
Evaluate the appearance of the adjusted image by each solution vector
To do. The adjustment history recording unit 3805 is the user evaluation determination unit 32.
The solution vector corresponding to the evaluation value obtained in 08 is recorded.
It The user adjustment end determination unit 4002 performs the adjustment now.
Number of iterations for a user who has a loop
Number of times loop^thJudge whether the condition has been exceeded. This
If the condition of is not satisfied, the goodness-of-fit calculation unit 105,
A new solution vector is obtained through the processing in the replacement operation unit 108.
Generate a rule set. Then, again, the second user evaluation unit 38
Return to 02 and evaluate each solution vector by the user.
U On the contrary, the end condition in the user adjustment end determination unit 4002
If the above condition is satisfied, the common model estimation determination unit 4004
A plurality of data recorded in the adjustment history recording unit 3805
Common model estimation condition with user adjustment history set in advance
It is determined whether or not is satisfied. This common model estimation condition
Can have various conditions, but here you can
To grasp the characteristics of the adjustment process common to many users
For the purpose, the number of users used by the adjustment process is set
Estimate whether common model exceeds the number of users
Selected for the conditions.

Here, if the number of users who have completed the adjustment is small, the process returns to the initial solution set setting unit 102, and the work of adjusting the vision correction lens by another user is started. If the number of adjustment use users exceeds the allowable number of adjustment use users, the common model estimation execution unit 4005 actually estimates the common model, assuming that there is enough data to estimate the common model of a plurality of users. Done.
As in the case of the optimization adjusting method and the optimization adjusting apparatus in the fourth example related to the present invention, there are many methods for estimating the common model. Here, the common model is also used by using the neural network. It will be estimated. FIG. 77 is a conceptual diagram of the neural network used for estimating the common model regarding the adjustment process of a plurality of users in this example. Here, rather than the model estimation based on the pattern classification by the hierarchical neural network by the backpropagation method used in the optimization adjusting method and the optimization adjusting apparatus in the fourth example related to the present invention, a relatively simple learning algorithm is used. We applied a neural network based on a learning vector quantization method that can classify patterns with at least a high number of samples. However, as in the case of the fourth example related to the present invention, application of a hierarchical neural network by the backpropagation method is also conceivable.

The neural network in Example 5 is
As shown in FIG. 77, this is a two-layer structure having six input layer neurons and n output neuron numbers. The number of input neurons 6 is the number of components of the solution vector p _k to be evaluated now, and the number of output neurons n corresponds to the rank of p _k in the solution vector set having n elements. This means that the input solution vector is 1
It indicates that it is classified into rank categories from to n.

The solution vector p (p₁, p₂, ..., p_n) Enter vector
Corresponding to the m-th rank category in the solution vector set
Input vector p^m(m = 1, ..., n), and
Number W _ij(i, j = 1,2, ..., n).

In this common model estimation execution unit 4005,
The neural network learns the adjustment histories of multiple users to estimate the common model. The neural network of this example uses n vectors W _i (W _i1 , W _i2 , ..., W _in ) that form the coupling coefficient W _ij to calculate n spaces of the solution vector p as an input vector. It works by dividing into areas. The n coupling coefficient vectors are called reference vectors, and one reference vector is associated with each area. The reference vector corresponds to the closest vector to all solution vectors included in the corresponding area. Where the input vector in the vector W _i
If the vector closest to p is vector W _c ,
(Equation 22), and the output u _i from the output layer neuron i becomes (Equation 23).

[0163]

[Equation 22]

[0164]

[Equation 23]

Learning is performed by updating only this vector W _c , and this update amount ΔW _c is executed according to the following (Equation 24).

[0166]

[Equation 24]

Η (n ^loop ) is a learning coefficient (0 <η (n ^loop ) <1) that monotonically decreases according to the learning count n ^loop . (Equation 24) is
The reference vector W _c indicates that the region boundary surface is formed by approaching the input vector p ^m when correctly classified and moving away from p ^m when not classified. Thus, in the learning process, the solution vector to be actually classified is used as the input learning vector, and the input learning vector is given in a sufficiently large amount (learning vector quantization method).
In this way, the common model estimation execution unit 4005 uses the history of adjustments by many users recorded in the adjustment history recording unit 3805 to perform learning of the neural network of this example, for example, and input the solution vector solution vector set. It is processed so that it can be classified into 1 to n rank categories. Then, the common model output unit 4001 outputs the coupling coefficient W _{ij of the} neural network as a parameter representing the estimated common model. Through the above processing, the fifth optimization adjusting method and the optimization adjusting apparatus related to the present invention relate to the adjustment process from the history of the solution vector optimization performed by a plurality of users using the interactive genetic algorithm. The common model is estimated, and common factors of the tastes of a plurality of users can be extracted.

The optimization adjusting method and the optimization adjusting apparatus in the sixth example related to the present invention will be described below. The optimization adjusting method and the optimization adjusting apparatus of the sixth example perform adjustment of a solution vector optimum for a user by using an interactive genetic algorithm, and update a common model regarding the adjustment process of a plurality of users. Is. Example 6
Similar to the first example related to the present invention, the adjustment problem of the lens for correcting the visual acuity is dealt with. FIG. 43 is a sixth diagram related to the present invention.
3 shows a configuration of an optimization adjusting device in the example of FIG. In FIG. 43, reference numeral 4301 denotes a common model evaluation calculation unit 4302 for calculating the evaluation value of each solution vector using a common model relating to the adjustment process of a plurality of users obtained before, and reference numeral 4302 denotes a common model evaluation calculation unit 4301 for the user. A common model evaluation determination unit that requests determination of whether the obtained evaluation value is appropriate,
A common model update unit 4303 updates the common model relating to the adjustment process of a plurality of users, which has been used so far, based on the adjustment history of the user recorded in the adjustment history recording unit 3805. The common model update unit 4303 determines whether or not the user's adjustment history recorded in the adjustment history recording unit 3805 satisfies a preset common model update condition, and a common model update determination unit 4304 and a common model update determination unit 4304. When the update condition is satisfied, the common model update execution unit 4305 is used to actually update the common model using the recorded user adjustment history.

The operation of the optimization adjusting method and the optimization adjusting apparatus in the sixth example related to the present invention configured as above will be described with reference to the flow charts of FIGS. 44 and 45.

First, the target data input unit 3201 is used.
The test image data to be shown to the user is input. Solution to adjust
Cutle p_kIs the optimization key in the first example related to the present invention.
This is the same as the case of the adjusting method and the optimization adjusting device. That is,
The solution vector is p_k= (L^SPH _k, L^{CY L} _k, L^AXIS _k, R^SPH _k,
R^CYL _k, R^AXIS _k). Initial solution set setting unit 1
01, the initial set of solution vectors P = [p_k] (k = 1, ...,
n) is set. Common model evaluation calculator 4301
For example, the optimization adjustment method of the fifth example related to the present invention
And the over-adjustment of multiple users obtained by using the
Each solution vector p using a common model_kEvaluation of
Value E_kTo calculate. Then, in the information presenting section 3207,
Visible through the corrective lens created by each solution vector
The adjusted image is presented to the user and the user can
Whether the evaluation value in the general model evaluation calculation unit 4301 is appropriate
It is the common model evaluation determination unit 4302 that determines
It If it is determined to be appropriate, then the conformity meter
Of the solution vector by the calculation unit 105 and the recombination operation unit 108
A new solution vector set is generated by the recombination operation.
It At that time, the number of repetitions exceeded the allowable number of repetitions.
Or the information you ask the user to rate
Satisfies the end condition of whether something good for the foot was created
In this case, the optimum solution output unit 103 and the device adjustment execution unit 3204
And asks if you want to adjust the vision correction lens.
Finish processing such as pressing. Do not meet this termination condition
If not, it returns to the common model evaluation calculation unit 4301 again.
And solve the calculation of the evaluation value of the new solution vector.
It

On the other hand, when the user determines that the evaluation value based on the common model is not appropriate by the common model evaluation determination unit 4302, the process proceeds to the second user evaluation unit 3802, and the user himself / herself evaluates each solution vector. And the evaluation value and the solution vector are recorded. As a common model update condition, the common model update determination unit 430 considers whether or not the history of adjustments newly performed by the user recorded in the adjustment history recording unit 3805 exceeds the allowable number of adjustment history.
4 does. If the newly recorded adjustment history is small, that is, if the common model update conditions are not satisfied,
The common model is not updated as it is, and a new solution vector set is generated by deriving the goodness of fit and recombining operations. On the other hand, when the common model update condition is satisfied, the common model update execution unit 4305 based on the adjustment history recorded in the adjustment history recording unit 3805 creates the previously estimated common model regarding the adjustment process of the plurality of users. Actually update. There are various methods as the method, but the model estimation method by the neural network described in the examples of the fourth and fifth optimization adjusting methods and the optimization adjusting apparatus related to the present invention is also conceivable. Therefore, this method is used also in this example, and the configuration of the neural network, the learning method, the teacher signal, etc. are omitted.

The new common model obtained by the common model update execution unit is immediately replaced with the conventional common model and used by the common model evaluation calculation unit 4301. By performing the processing as described above, the sixth aspect related to the present invention
The optimization adjustment method and the optimization adjustment apparatus in the example of (1) perform the adjustment of the solution vector that is optimum for the user, and also work to update the common model regarding the adjustment process of a plurality of users, so that the user can adjust all solution vectors. It is possible to reduce the burden on the user without having to evaluate the information according to.

The optimization adjusting method and the optimization adjusting apparatus in the seventh example related to the present invention will be described below with reference to the drawings. The seventh optimization adjusting method and the optimization adjusting apparatus estimate an evaluation model regarding a user's adjustment process using an interactive genetic algorithm, and automatically use the obtained evaluation model to perform optimum adjustment of a solution vector. Is to do. The seventh embodiment deals with the same problem as the first embodiment related to the present invention. FIG. 46 shows a seventh embodiment related to the present invention.
5 shows a configuration of an optimization adjustment device in the embodiment. In FIG. 46, reference numeral 4601 is a method selection switch for selecting whether to perform evaluation by the user or evaluation by the evaluation model, and reference numeral 4602 indicates an evaluation of the adjustment process of the user based on the adjustment history of the user recorded in the adjustment history recording unit 3805. An evaluation model estimation unit that estimates the model, and a model evaluation calculation unit 4603 that calculates the evaluation value of each solution vector using the obtained evaluation model regarding the adjustment process of the user. Then, the evaluation model estimation unit 4602 uses the adjustment history recording unit 3
It is determined whether or not the user's adjustment history recorded in 805 satisfies a preset evaluation model estimation condition. The model estimation execution unit 3804 is included.

The operation of the optimization adjusting method and the optimization adjusting apparatus in the seventh example related to the present invention configured as above will be described with reference to the flow charts of FIGS. 47 and 48.

First, the target data input section 3201 is used.
The test image data to be shown to the user is input. Solution to adjust
Cutle p_kIn the first embodiment related to the present invention
This is the same as the case of the optimization adjustment method and the optimization adjustment device.
That is, the solution vector is p_k= (L^{S PH} _k, L^CYL _k, L^AXIS _k, R^SPH
_k, R^CYL _k, R^AXIS _k). Initial solution set setting
In the unit 101, an initial solution set P = [p of solution vectors_k] (k =
1, ..., n) is set. Method selection switch allows the user
Select whether to perform evaluation using the evaluation model or the evaluation model.
Selection is made. At the start of the process, the second user evaluation unit 38
02 is selected. In the information presenting unit 3207, each solution
Adjusted image seen through a corrective lens made by Tor
Is presented to the user, and the user evaluation determination unit 3208 displays it.
Based on this, each solution vector is evaluated. Pair with this evaluation value
The corresponding solution vector is recorded in the adjustment history recording unit 3805.
It The history recorded in the adjustment history recording part 3805 is
Switching the method when the evaluation model estimation conditions set for
Model estimation execution unit 3 when the determination unit 4604 makes a determination.
804 estimates an evaluation model for the user's adjustment process
It At the same time, the method switching determination unit 4604
The method selection switch 4601 is instructed to switch the evaluation method.
It After receiving this instruction, the evaluation by the user
2 instead of the user evaluation unit 3802, the model evaluation calculation unit 4
In 603, obtained by the model estimation execution unit 3804
Each solution vector is calculated using the evaluation model for the user's adjustment process.
Tolu is evaluated. Evaluation model estimation conditions, model estimation
The constant execution unit 3804 has a fourth function related to the present invention.
An optimization adjusting method and an optimization adjusting device according to an embodiment
Since it is similar, it is omitted. The model evaluation calculation unit 4603 is also
Specifically, each solution vector obtained by the user evaluation judgment unit 3208
Based on the evaluation value of the
Of the relevance calculation unit 1 based on the calculated relevance
08 performs arithmetic recombination operation of each solution vector.

The above-described processing satisfies the preset termination condition that the number of repetitions exceeds the allowable number of repetitions, or that the information requested to be evaluated by the user is created enough to satisfy the user. The optimum solution vector is obtained by repeating the procedure up to. Then, the optimum solution vector is output by the optimum solution output unit 103, and the device adjustment execution unit 3204 performs processing such as executing adjustment of the vision correction lens or issuing an instruction thereof. Through such a series of processing, it becomes possible to estimate the evaluation model related to the user's adjustment process using the interactive genetic algorithm and automatically adjust the optimal solution vector using the obtained evaluation model. Thus, it is possible to reduce the burden on the user, which has been a problem in putting the interactive genetic algorithm into practical use, and efficiently realize a device optimally adjusted to the user's taste.

Furthermore, in the optimization adjusting method and the optimization adjusting apparatus in the first to seventh examples related to the present invention,
An example in which it is applied to the problem of creating a font of characters according to the user's taste can be considered. Hereinafter, the case where the optimization adjusting methods and the optimization adjusting devices of the first to seventh examples of the present invention are applied to the creation of the personal character font will be described.

As a character font, the outline of a character is defined as a curve in a virtual coordinate space, and x, y coordinate data of a node s _i (i = 1, ..., m) selected from the curve. The outline font expressed by is often used. For example, when the character "b" is represented by an outline font, it is represented by the coordinates of 13 black circle nodes s _i . Then, from these coordinates of the nodes, the contour lines are interpolated by a curve formula. A quadratic or cubic spline function, a cubic Bezier curve, or the like is used as the curve formula, but here, the curve approximation between the nodes is performed using the cubic function of (Equation 25).

[0179]

[Equation 25]

In (Equation 25), (X _a , Y _a ), (X _b , Y _b ) represent the coordinates of the nodes a and b of the character expressed in outline font, and (X, Y) represent the node a, The coordinates of the interpolation points between b are shown. Further, α, β, γ, and δ represent coefficients that take arbitrary real values. The expressions α, β, γ in (Equation 25) are necessary conditions for the approximated curve to pass through both the nodes a and b. In consideration of these points, the solution vector p _k for generating the character font is the x coordinate X _si , y coordinate Y of the node s _i (i = 1, ..., m) previously obtained by the outline font representation. _si and nodes s _i and s
It is configured by arranging the coefficients α _i , β _i , γ _i , and δ _i of the approximation curve between _{i + 1 in} the order of nodes. Data of the test character group expressed in outline font is input to the target data input unit 3201. The information presenting unit 3207 creates an adjusted character group from the test character group represented by the outline font using each solution vector and presents it on the screen, and the user evaluation determining unit 3208 determines the evaluation value of each solution vector. To get it. In the case of the optimization adjusting method and the optimization adjusting apparatus of the first example related to the present invention, the update area limiting unit 3202 for determining the limited area that each solution vector can take is provided, but here the node s _i i
The x-coordinate X _si and the y-coordinate Y _si of (1, ..., M) are limited to the area Λ represented by ( _Equation 26).

[0181]

[Equation 26]

Here, (x _i0 , y _i0 ) represents the coordinate data of the node of the test character group expressed by the outline font first input to the target data input unit 3201, and r _i can be taken by each node. It represents the radius of the circular region Λ. This is to limit the movable area of each node in order to exclude the case where the outline font characters actually obtained from the solution vector are displayed as completely different characters or images when displayed on the screen. It was done. Further, after the solution vector having the highest evaluation value is output as the optimum solution vector by the optimum solution output unit 103, characters can be displayed using the outline font personally created by the device adjustment execution unit 3204 according to the optimum solution vector. As described above, the character output means is adjusted or instructed. In the adjusting devices of the first to seventh examples related to the present invention, the remaining components operate in the same manner as in the above-described first to seventh examples, and thus the description thereof will be omitted. However, even in the present example in which the optimization adjusting methods and the optimization adjusting devices of the first to seventh examples related to the present invention are applied to this problem, the character font based on the user's preference can be obtained without any special knowledge. By making it possible to delete the search in the area unnecessary for the search, create an evaluation model of the adjustment process from the user's adjustment history, and perform automatic adjustment using the model. , It is possible to derive an optimal solution vector that can efficiently create a character font most liked by the user. As a result, it is considered that it is possible to reduce the burden on the user who performs the evaluation, which was a problem in the conventional technique.

The optimization adjusting method and the optimization adjusting apparatus in the eighth example relating to the present invention will be described below. The optimization adjusting method and the optimization adjusting apparatus of the eighth example interact the information for assisting in comparison / evaluation of each solution vector and the function of rearranging the order of presentation in a problem dealing with time series signals. The optimal solution vector is adjusted by adding it to the type genetic algorithm. Example 8 is an example applied to the problem of improving the sound quality of a voice distorted due to information loss during transmission. FIG. 49 shows the structure of the adjusting device in the eighth example related to the present invention.

In FIG. 49, reference numeral 4901 denotes a time-series evaluation execution unit for presenting the time-series information represented by each solution vector to the user for evaluation by the user. Time series evaluation execution unit 49
Reference numeral 01 has a role of helping the memory of the user when the information presenting unit 3207 that presents the time-series information represented by each solution vector to the user and the plurality of time-series information presented by the information presenting unit 3207 are compared and evaluated. Auxiliary information presenting section 4902 for presenting auxiliary information, user evaluation judging section 3208 for having the user actually evaluate each solution vector, and each time series information based on the evaluation value obtained by user evaluation judging section 32084 And information rearranging section 4 for changing the order of presenting auxiliary information
903.

The operation of the optimization adjusting method and the optimization adjusting apparatus in the eighth example related to the present invention configured as described above will be described with reference to the flowchart of FIG. 50, which is shown in FIG. 78 here. An example applied to the sound quality improvement problem of distorted speech performed in the procedure described below will be described. The purpose of this problem is to ask the user to hear the distorted voice and adjust the coefficient of the distortion improving filter so as to improve the sound quality based on the hearing condition of the sound. First, as shown in FIG. 78, in the target data input unit 3201,
Distorted voice is input due to information loss during transmission. There are various ways to improve the sound quality of distorted speech, but in this example, as shown in the figure, FIR filter (finite interval impulse response filter: Finite Impulse
Response filter) is used as a distortion improving filter to improve the sound quality of distorted speech. Then, the filter coefficient vector a = (a ₀ , a ₁ , ..., A _step ) of the FIR filter having a step number of steps is adjusted by the interactive genetic algorithm. Therefore, the solution vector p _{k to be} adjusted is defined by arranging the filter coefficients a _i . In the solution vector, each filter coefficient is handled as a real value as in the above examples, but it is also possible to convert a _i into, for example, a binary number of Blen and to handle the binary number side by side. The initial solution set setting unit 101 sets an initial set of solution vectors P = [ _pk ] (k = 1, ..., n) according to this definition. Information presenting section 32
At 07, a distortion improving filter configured by each solution vector is prepared, and the distortion improving sound obtained by filtering the distorted sound input by the target data input unit 3201 is presented to the user.

However, even if n distortion-improved speeches are presented as they are, it becomes more difficult to compare and evaluate the differences between the speeches as the number of n becomes larger. Since static data such as images and graphics can be spatially arranged on the display at the same time, it is easy to compare and evaluate. On the other hand, when dealing with voices, it is possible to compare and evaluate two voices, but if the number of voices exceeds that, it becomes very difficult to compare and evaluate. This is because the user is confused by listening to too many voices and cannot distinguish the difference between the voices. Therefore, the auxiliary information presenting section 4902 presents auxiliary information for facilitating the comparative evaluation of each voice. Although various kinds of information can be considered as the auxiliary information to be presented, in this example, a face image obtained by converting the filter coefficient into the parameters of the size, angle, and position of the eyes, mouth, and nose forming the face ( (Figure 79). Although the face itself has nothing to do with the sound quality of the speech after distortion improvement, we think that it will help the user to remember when simultaneously comparing and evaluating n distortion-corrected speeches. Based on these pieces of information, the user evaluation determination unit 3208 determines the evaluation value of each solution vector by the user. Further, the information rearranging section 4903 rearranges the distortion-improving voice to be compared with the face image which is its auxiliary information based on each evaluation value determined by the user evaluation determining section 3208. This allows the user to visually rearrange the superiority or inferiority of each distortion-improved voice, which leads to a reduction in the influence of fluctuations in the user's evaluation. The process in the time-series evaluation execution unit 4901 is repeatedly performed until the evaluation by the user is completed.

From the evaluation values thus obtained, the goodness-of-fit calculation unit 105 calculates the goodness-of-fit of each solution vector, and the recombining operation unit 108 recomposes each solution vector based on the goodness-of-fit. The above process is repeated lo
The operation is performed until op becomes larger than the allowable repeat count loop ^th or the user is satisfied with the distortion improving voice, and the device adjustment execution unit 3
The filter represented by the optimum solution vector obtained in 204 is configured, and the sound quality of distorted speech is improved. As described above, in the problem of dealing with time-series signals, the optimal solution can be obtained by adding the information that assists in comparing and evaluating each solution vector and the function of rearranging the order of presentation to the interactive genetic algorithm. By adjusting the vector, it becomes possible to apply the interactive genetic algorithm, which has been applied only to static data such as images, to dynamic data such as time series data. As a result, for example, even in the case of adjustment according to the hearing characteristics of the person who wears the hearing aid, it is not necessary to bother to go to an adjuster who has specialized knowledge, and the deaf person who is the user can easily make the adjustment. In the past, only the volume could be adjusted, but by using the adjusting device of this example, even if the hearing condition changes, it is possible to immediately adjust the sound quality according to one's own hearing condition.

The optimization adjusting method and the optimization adjusting apparatus in the ninth example related to the present invention will be described below. The ninth optimization adjustment method and the optimization adjustment device have a function of rearranging the order of presentation and information to assist in comparison / evaluation of each solution vector in a problem dealing with time series signals. The search area is limited based on its characteristics and past adjustment results, and the optimal solution vector is adjusted by an interactive genetic algorithm. In Example 9, Example 8
As in the case of, we deal with the problem of creating a filter that improves the quality of distorted speech. FIG. 51 shows the configuration of the optimization adjusting apparatus in the ninth example related to the present invention.

As is apparent from FIG. 51, the optimization adjusting apparatus of Example 9 is the same as the optimization adjusting apparatus of the first example related to the present invention and the optimization adjusting apparatus of the eighth example related to the present invention. It has a configuration that includes a characteristic time-series evaluation execution unit.

The operation of the optimization adjusting apparatus in the ninth example related to the present invention follows the flowchart of FIG. 52, and when compared with the flowchart of the optimization adjusting apparatus in the first example related to the present invention, the target data Input unit 32
In 01, a sound distorted due to information loss at the time of transmission is input, the solution vector p _k to be adjusted is defined by arranging the coefficients a _i of the distortion improving filter, and in the information presenting unit 3207, each solution vector is Presenting a distortion-improved voice obtained by performing a filtering process on the distorted sound input by the target data input unit 3201 by providing a distortion-improved filter configured by the auxiliary data presenting unit 49.
02, presenting a face image obtained by converting the filter coefficient into parameters of the size, angle, and position of the eyes, mouth, and nose forming the face as auxiliary information for facilitating comparative evaluation of each voice. The information rearranging unit 4903 rearranges the distortion-improved voice to be compared with each other based on each evaluation value determined by the user evaluation determination unit 3208 and the face image which is the auxiliary information thereof, and the time-series evaluation execution unit. 49
The processing in 01 is repeated until the evaluation by the user is completed, and the filter represented by the optimum solution vector obtained by the device adjustment execution unit 3204 is configured, and the improvement of the sound quality of distorted speech is different. Only. The rest of the components and the flow of processing are the same as in the case of the optimization adjusting method and the adjusting device in the first example related to the present invention, and therefore will be omitted.

However, as shown in the flow chart of FIG. 52, information to assist in comparing / evaluating solution vectors and a function of rearranging the order of presentation are added to the interactive genetic algorithm, and the solution vector to be adjusted is adjusted. By limiting the area that can be taken by and saving the effort of the search processing in an unnecessary area, it is possible to efficiently perform dynamic data adjustment such as time series information using an interactive genetic algorithm. As described above, the optimization adjusting method and the optimization adjusting apparatus in the ninth example related to the present invention are the first and the eighth.
It has an effect that combines the effects of the optimization adjustment method and the optimization adjustment device in the above example.

The optimization adjusting method and the optimization adjusting apparatus in the tenth example related to the present invention will be described below.
The tenth optimization adjusting method and the optimization adjusting device have a function of rearranging the order of presenting information and information for assisting in comparison / evaluation of each solution vector in a problem dealing with a time-series signal, and recording The initial set of solution vectors is set based on the past adjustment information, and the solution vectors are optimized using an interactive genetic algorithm. Example 10
Now, as in the case of Example 8, the filter creation problem for improving the sound quality of distorted speech is treated. FIG. 53 is a tenth view related to the present invention.
3 shows a configuration of an optimization adjusting device in the example of FIG.

As can be seen from FIG. 53, the optimization adjusting apparatus of Example 10 is the same as the optimization adjusting apparatus of the first example related to the present invention and the optimization adjusting apparatus of the eighth example related to the present invention. It has a configuration in which a time-series evaluation execution unit, which is a feature of, is added.

The optimization adjusting apparatus in the tenth example related to the present invention operates according to the flowchart of FIG. 54, and is a flowchart showing the processing steps of the optimization adjusting apparatus in the first example related to the present invention. By comparison,
In the target data input unit 3201, a voice that is distorted due to information loss during transmission is input, and the solution vector p _k to be adjusted is defined by arranging the coefficients a _i of the distortion improving filter. , Presenting a distortion-improved voice obtained by preparing a distortion-improvement filter composed of each solution vector and filtering the distorted sound input by the target data input unit 3201, and from the auxiliary information presenting unit 4902. Presenting a face image obtained by converting the filter coefficient into parameters of eyes, mouth, nose size, angle, and position that form the face as auxiliary information for facilitating comparative evaluation of voice, In the replacement unit 4903, the user evaluation determination unit 320
Rearrangement of the distortion-improved voice to be compared based on each evaluation value determined in 8 and the face image as its auxiliary information,
The processing in the time-series evaluation execution unit 4901 is repeatedly performed until the evaluation by the user is completed, and the filter represented by the optimum solution vector obtained by the device adjustment execution unit 3204 is configured to improve the sound quality of distorted speech. Is only different. The rest of the components and the flow of processing are the same as in the case of the adjusting device in the first example related to the present invention, and are therefore omitted.

However, as shown in the flowchart of FIG. 54, by adding to the interactive genetic algorithm a function for rearranging the order of information and information for assisting in comparison / evaluation of solution vectors, user evaluation Can reduce the burden on. In addition, by optimizing the solution vector by using multiple past optimum solution vectors for the initial set that starts the search, it is possible to efficiently and interactively adjust dynamic data such as time series information. A statistical algorithm can be applied. Thus, the first related to the present invention
The optimization adjustment method and the optimization adjustment device in the example of 0 are the first
And the effects of the optimization adjusting method and the optimization adjusting device in the eighth example are combined.

The optimization adjusting method and the optimization adjusting apparatus in the eleventh example related to the present invention will be described below.
The eleventh optimization adjustment method and the optimization adjustment device have a function of rearranging the order of presentation and information for assisting in comparison / evaluation of each solution vector in a problem dealing with a time series signal. The user's evaluation value is corrected based on the psychological condition estimated based on the physiological information of the user, and the optimal solution vector is derived using the interactive genetic algorithm based on the correction value of the evaluation value. Is. Example 11 deals with the problem of filter creation for improving the sound quality of distorted speech as in Example 8. FIG. 55 shows the configuration of the optimization adjusting apparatus in the eleventh example related to the present invention.

As can be seen from FIG. 55, the optimization adjusting apparatus of Example 11 is the same as the optimization adjusting apparatus of the third example related to the present invention and the optimization adjusting apparatus of the eighth example related to the present invention. It has a configuration in which a time-series evaluation execution unit, which is a feature of, is added.

The operation of the optimization adjusting apparatus in the eleventh example related to the present invention follows the flowchart of FIG. 56, and when compared with the flowchart of the optimization adjusting apparatus in the third example related to the present invention, the target data Input unit 32
In 01, a sound distorted due to information loss at the time of transmission is input, the solution vector p _k to be adjusted is defined by arranging the coefficients a _i of the distortion improving filter, and in the information presenting unit 3207, each solution vector is Presenting a distortion-improved voice obtained by performing a filtering process on the distorted sound input by the target data input unit 3201 by providing a distortion-improved filter configured by the auxiliary data presenting unit 49.
02, presenting a face image obtained by converting the filter coefficient into parameters of the size, angle, and position of the eyes, mouth, and nose forming the face as auxiliary information for facilitating comparative evaluation of each voice. The information rearranging unit 4903 rearranges the distortion-improved voice to be compared with each other based on each evaluation value determined by the user evaluation determination unit 3208 and the face image which is the auxiliary information thereof, and the time-series evaluation execution unit. 49
The processing in 01 is repeated until the evaluation by the user is completed, and the filter represented by the optimum solution vector obtained by the device adjustment execution unit 3204 is configured, and the improvement of the sound quality of distorted speech is different. Only. The rest of the components and the flow of processing are the same as in the case of the optimization adjusting method and the optimization adjusting apparatus in the third example related to the present invention, and therefore will be omitted.

However, as shown in the flowchart of FIG. 56, the user can add the information for assisting in comparing / evaluating solution vectors and the function of rearranging the order of presentation to the interactive genetic algorithm so that the user can The load on the evaluation of the solution vector is reduced, and the psychological situation of the user who is performing the estimation is estimated and its effect is corrected, so fluctuations in the evaluation due to the psychological situation of the user (excitement, indifference, etc.) are suppressed as much as possible. It is possible to optimize the solution vector. As described above, the optimization adjusting method and the optimization adjusting device in the eleventh example related to the present invention are the third and eighth embodiments.
It has an effect that combines the effects of the examples.

The optimization adjusting method and the optimization adjusting apparatus in the twelfth example related to the present invention will be described below.
The twelfth optimization adjustment method and the optimization adjustment device provide presentation of information to assist in comparison and evaluation of solution vectors and rearrangement of presentation order by evaluation values in problems dealing with time-series signals. The evaluation model for the user's adjustment process is estimated based on the adjustment history of the user by devising and using the interactive genetic algorithm.
The example 12 deals with the filter creation problem for improving the sound quality of the distorted voice as in the case of the example 8. FIG. 57 shows the configuration of the optimization adjusting apparatus in the twelfth example related to the present invention.

As can be seen from FIG. 57, the optimization adjusting device of Example 12 is the same as the adjusting device of the fourth example relating to the present invention and the characteristics of the optimization adjusting device of the eighth example relating to the present invention. It has a configuration in which a time series evaluation execution unit is added.

The optimization adjusting apparatus in the twelfth example related to the present invention operates according to the flowchart of FIG. 58, and is compared with the flowchart of the optimization adjusting apparatus in the fourth example related to the present invention. Then, in the target data input unit 3201, a voice distorted due to information loss at the time of transmission is input, and the solution vector p _k to be adjusted.
Is defined by arranging the coefficients a _i of the distortion improving filter. The information presenting unit 3207 prepares a distortion improving filter composed of each solution vector, and the target data input unit 32
The distortion-improved sound obtained by filtering the distorted sound input at 01 is presented to the user, and the auxiliary information presenting unit 4902 sets the filter coefficient as the auxiliary information for facilitating the comparative evaluation of each face. Presenting a face image obtained by converting into parameters of the size, angle, and position of the constituent eyes, mouth, and nose. In the information rearranging unit 4903, each evaluation value determined by the user evaluation determining unit 3208 is displayed. The distortion-improved voice to be compared with the original and rearrangement of the face image that is the auxiliary information are rearranged, the processing in the time-series evaluation execution unit 4901 is repeatedly performed until the evaluation by the user is completed, and the device adjustment execution unit 32
The only difference is that the filter represented by the optimum solution vector obtained in 04 is configured and the improvement of the sound quality of distorted speech is realized. For the remaining components and process flow,
Since it is the same as the case of the optimization adjusting method and the optimization adjusting apparatus in the fourth example related to the present invention, description thereof will be omitted.

However, as shown in the flow chart of FIG. 58, an interactive genetic algorithm is devised to present information to assist in comparison and evaluation of each solution vector and rearrange the order of presentation according to evaluation values. In addition to being used, the evaluation model for the user's adjustment process is estimated based on the user's adjustment history, and the user's preference, which has been difficult until now, can be known even in the time-series information adjustment. As described above, the optimization adjusting method and the optimization adjusting apparatus in the twelfth example related to the present invention also have the functions which are expanded so that the optimization adjusting apparatus in the fourth example can be applied to the time series signal problem. -ing

The optimization adjusting method and the optimization adjusting apparatus in the thirteenth example related to the present invention will be described below.
The thirteenth optimization adjusting method and the optimization adjusting device provide information for assisting in comparison / evaluation of solution vectors and rearrangement of the order of presentation by evaluation values in a problem dealing with time-series signals. By devising and using the interactive genetic algorithm, a common model for the adjustment process of a plurality of users is estimated based on the recorded adjustment history of a plurality of users. The example 21 deals with the filter creation problem for improving the sound quality of the distorted voice as in the case of the example 16. FIG. 59 shows the configuration of the optimization adjusting apparatus in the thirteenth example related to the present invention.

As can be seen from FIG. 59, the optimization adjusting apparatus of Example 13 is the same as the optimization adjusting apparatus of the fifth example related to the present invention and the optimization adjusting apparatus of the eighth example related to the present invention. It has a configuration in which a time-series evaluation execution unit, which is a feature of, is added.

The optimization adjusting apparatus in the thirteenth example related to the present invention operates according to the flow charts of FIGS. 60 and 61, and is the same as the optimization adjusting apparatus in the fifth example related to the present invention. Compared with the flowchart, the target data input unit 3201 inputs a sound distorted due to information loss at the time of transmission, and the solution vector p _k to be adjusted is defined by arranging the coefficients a _i of the distortion improving filter. The information presenting unit 3207 presents a distortion improving voice obtained by preparing a distortion improving filter composed of each solution vector and filtering the distorted sound input by the target data input unit 3201 to the user. From the information presenting unit 4902, filter coefficients are used as auxiliary information for facilitating comparative evaluation of each voice, eyes, mouth, The size of the angle, to present a face image obtained by converting the parameters of position,
In the information rearranging section 4903, the user evaluation judging section 3
The distortion-improved voice to be compared based on each evaluation value determined in 208 and the face image which is its auxiliary information are rearranged, and the processing in the time-series evaluation execution unit 4901 is performed until the evaluation by the user is completed. The only difference is that the filter represented by the optimum solution vector obtained by the device adjustment execution unit 3204 is configured repeatedly and the improvement of the sound quality of the distorted voice is realized. The remaining components and the flow of processing are the same as in the case of the optimization adjusting method and the optimization adjusting apparatus in the fifth example related to the present invention, and therefore will be omitted.

However, as shown in the flow charts of FIGS. 60 and 61, the interactive genetic algorithm is used to present information to assist in comparison and evaluation of each solution vector and rearrange the order of presentation according to evaluation values. By devising and using it, by estimating the common model for the adjustment process of multiple users based on the recorded adjustment history of multiple users,
It is possible to extract common factors of preference when handling time-series signals by a plurality of users, which was difficult only with conventional interactive genetic algorithms.

The optimization adjusting method and the optimization adjusting apparatus in the fourteenth example related to the present invention will be described below.
The fourteenth optimization adjustment method and the optimization adjustment device provide information for assisting in comparison and evaluation of solution vectors and rearrangement of the order of presentation by evaluation values in problems dealing with time-series signals. It has a function to perform and derives an optimal solution vector for a user by using an interactive genetic algorithm, and updates a common model regarding the adjustment process of a plurality of users. In Example 14, as in Example 8, the filter creation problem for improving the sound quality of distorted speech is dealt with. FIG. 62 shows the configuration of the optimization adjusting apparatus in the fourteenth example related to the present invention.

As can be seen from FIG. 62, the optimization adjusting apparatus of Example 14 is the same as the optimization adjusting apparatus of the sixth example related to the present invention and the optimization adjusting apparatus of the eighth example related to the present invention. It has a configuration in which a time-series evaluation execution unit, which is a feature of, is added.

The operation of the adjusting device in the fourteenth example related to the present invention follows the flowcharts of FIGS. 63 and 64, and is compared with the flowchart of the optimizing adjusting device in the sixth example related to the present invention. In the target data input unit 3201, a voice distorted due to information loss at the time of transmission is input, the solution vector p _k to be adjusted is defined by arranging the coefficients a _i of the distortion improving filter, and the information presenting unit 3207. Then, a distortion improvement filter composed of each solution vector is prepared, and the target data input unit 3201
Presenting the distortion-improved voice obtained by filtering the distorted voice input by the user, and constructing a filter coefficient face as auxiliary information for facilitating comparative evaluation of each voice from the auxiliary information presenting unit 4902. Eye,
Presenting a face image obtained by converting into parameters of mouth, nose size, angle, and position, and comparing information based on each evaluation value determined by the user evaluation determination unit 3208 in the information rearranging unit 4903. The distortion-correcting voice and the face image which is the auxiliary information thereof are rearranged, the processing in the time-series evaluation executing unit 4901 is repeatedly executed until the evaluation by the user is completed, and the device adjustment executing unit 3204.
The only difference is that the filter represented by the optimum solution vector obtained in step 1 is constructed and the improvement of the sound quality of distorted speech is realized. The remaining components and the flow of processing are the same as in the case of the optimization adjusting method and the optimization adjusting apparatus in the sixth example related to the present invention, and are therefore omitted.

However, as shown in the flow charts of FIGS. 63 and 64, interactive genetic inheritance is achieved by the function of presenting information to assist in comparison and evaluation of solution vectors and rearranging the order of presentation according to evaluation values. It is possible to reduce the burden on the user by applying the dynamic algorithm to the adjustment of the time series information, which is conventionally unsuitable, and by working to update the common model regarding the adjustment process of a plurality of users.

The optimization adjusting method and the optimization adjusting apparatus in the fifteenth example related to the present invention will be described below.
The fifteenth optimization adjusting method and the optimization adjusting device provide information for assisting comparison and evaluation of solution vectors and rearrangement of the order of presentation by evaluation values in a problem dealing with a time-series signal. It has a function to perform, estimates an evaluation model for the user's adjustment process using an interactive genetic algorithm, and automatically adjusts the optimal solution vector using the obtained evaluation model. The example 15 deals with the filter creation problem for improving the sound quality of the distorted voice as in the case of the example 8. FIG. 65 shows the configuration of the optimization adjusting apparatus in the fifteenth example related to the present invention.

As can be seen from FIG. 65, the optimization adjusting apparatus of Example 15 is the same as the optimization adjusting apparatus of the seventh example related to the present invention and the optimization adjusting apparatus of the eighth example related to the present invention. It has a configuration in which a time-series evaluation execution unit, which is a feature of, is added.

The operation of the optimization adjusting apparatus in the fifteenth example related to the present invention follows the flowcharts of FIGS. 48 and 66, and the flowchart of the optimization adjusting apparatus in the seventh example related to the present invention. Compared with the above, the target data input unit 3201 receives a sound that is distorted due to information loss during transmission, and a solution vector to be adjusted.
p _k is defined by arranging the coefficients a _i of the distortion improving filter, and the information presenting unit 3207 prepares a distortion improving filter composed of each solution vector, and the target data input unit 3
Presenting the distortion-improved voice obtained by filtering the distorted voice input in 201 to the user,
A face image obtained from the auxiliary information presenting unit 4902 by converting the filter coefficient into the parameters of the size, angle, and position of the eyes, mouth, and nose forming the face as auxiliary information for facilitating the comparative evaluation of each voice. The information rearranging unit 4903 rearranges the distortion-improved voice to be compared with the face image which is the auxiliary information thereof based on each evaluation value determined by the user evaluation determination unit 3208. The processing in the series evaluation execution unit 4901 is repeatedly performed until the evaluation by the user is completed, and the device adjustment execution unit 32
The filter represented by the optimum solution vector obtained in 04 is configured, and it is different in that the sound quality of distorted speech is improved. The remaining components and the flow of processing are the same as in the case of the optimization adjusting method and the optimization adjusting apparatus in the seventh example related to the present invention, and therefore will be omitted.

However, as shown in the flow charts of FIGS. 48 and 66, information that assists in comparing / evaluating solution vectors and a function of rearranging the order of presentation are added to the interactive genetic algorithm and By generating an evaluation model of the user's adjustment process using the history of the user's adjustments that have been made and automatically optimizing the solution vector using this evaluation model instead of the evaluation by the user, The interactive genetic algorithm, which was applied only to static data such as images, can be efficiently applied to the adjustment of dynamic data such as time series data, and the burden on the evaluation user is greatly reduced. It becomes possible to do.

Furthermore, the eighth to fifteenth aspects related to the present invention.
In the optimization adjustment method and the optimization adjustment apparatus in the example of FIG. 12, it is possible to consider an example in which the optimization adjustment method and the optimization adjustment apparatus are applied to the problem of the characteristic adjustment of the hearing aid that matches the hearing characteristic of the user who is a hearing-impaired person. Below, the eighth to the first related to the present invention
The case where the optimization adjustment method and the optimization adjustment device of No. 5 are applied to the adjustment matching the characteristics of the person who wears the hearing aid will be described.

Normally, the characteristic adjustment of the hearing aid is from 125 Hz to 800.
The strength L ^f of the minimum audible value at the frequency f (Hz) (the sound with the lowest power) when a pure tone is heard at 0 Hz
(dB), the maximum audible value (the loudest sound that can withstand) H ^f (dB), and the normally heard sound intensity (dB) M ^f based on the values at multiple frequencies. This is done by adjusting the gain of the hearing aid. 125Hz to 8000Hz
^Is divided into three, and the m ^ear points in each frequency domain
If two values are extracted, the parameter used for characteristic adjustment is 3 × m ^ear . In addition to this, (i) consonants have low energy and are masked by vowels that follow consonants, so it is necessary to make consonants audible even when masked by emphasizing the consonants. (ii) It is easier to hear when the low-frequency component (formant component), which greatly affects the sound quality of the human ears, is emphasized. By taking into account the above two points, a more natural voice can be realized by the hearing aid. In the case of (i), it can be controlled by the limit P ^limit (dB) of the degree of consonant emphasis and the release time t ^rel (msec) of the emphasis that determines how much the transition from the consonant to the vowel continues to be emphasized.
In the case of (ii), P ^sup (dB) that suppresses the valley with respect to the peak (formant) of the speech spectrum in the low frequency range,
It is considered that the frequency width f ^wid (Hz) can be controlled. So, these four parameters
The solution vector p is defined in addition to the parameters of 3 × m ^ear .
In the target data input unit 3201, the test voice is input. The information presenting unit 3207 presents to the user the adjusted voice obtained through the test voice input by the target data input unit 3201 to the hearing aid represented by each solution vector. In particular, there is a certain relationship between the parameters forming the solution vector, and it seems that the region that the parameters can take is limited. It is considered that the adjustment of the solution vector within the limited area, which is a feature of the optimization controller, has a great meaning. The optimum solution output unit 103 outputs the solution vector having the highest evaluation value as the optimum solution vector, and the characteristic adjustment of the hearing aid is performed by the device adjustment execution unit 3204 based on the solution vector.
Done in. Here, the auxiliary information presentation unit 4902 handles a face image obtained from these parameters, as in the case of Example 16, for example.

Considering the optimization adjusting method and the optimization adjusting device in the eighth to fifteenth examples related to the present invention which are set as described above, the same as in the case of the problem of creating a sound quality improvement filter for distorted speech. In addition, even if the user has no special knowledge, it is possible to create a hearing aid based on the hearing condition that matches the hearing characteristics of the user. Also, an optimization adjustment method and an optimization adjustment device of an evaluation model regarding a user's adjustment process in a twelfth example related to the present invention, and an optimization adjustment method and optimization of a common model regarding an adjustment process of a plurality of users in an thirteenth example. It is considered possible to clearly describe the relationship between each parameter by using the chemical adjustment device.

In the optimization adjustment methods and the optimization adjustment devices in the first to eighth embodiments and the first to fifteenth examples of the present invention, the roulette selection method is used as the selection selection of the solution vector. It is not limited to this, combining an elite strategy method that directly copies a solution vector set that has a high degree of goodness of fit and is newly created next to a solution vector, and does not use the value of the degree of goodness of fit and pay attention only to its rank. It is also possible to apply a linear normalization method or the like. Further, even in the case of evaluation by the user, the maximum possible evaluation value EA is used in the present invention.
A value between _max and the minimum evaluation value EA _min was considered as a continuous value, but a graded evaluation value based on the relative evaluation of n solution vectors presented to the user at the same time, for example, n evaluation values. It may be possible to use the rank of the solution vector as the evaluation value. Furthermore, when the solution vector is represented by a bit string code, even in the crossover process, 1 in this example is used.
Not only the point or two-point crossover process but also the simplex crossover process as shown in FIG. 74 can be applied. As shown in FIG. 74, in the simplex crossover process, (i) first, two solution vectors having a high fitness and one solution vector having a low fitness are selected according to the probabilities used in the selection probability part. (ii) Two solution vectors having a high degree of conformity are compared with each other, and if the corresponding bit string code values match, the value is adopted. If they do not match, a new solution vector is created by taking the negation of the value of the corresponding bit of the low-fitness solution vector.

The following processing is executed. The simplex crossover process has an advantage over the one-point or two-point crossover process in that the adjustment can be performed while maintaining the diversity of the solution vector, and the local solution is less likely to fall.

From the above, the optimization adjusting method and the optimization adjusting apparatus of the first example related to the present invention limit the update area of each parameter to be adjusted based on its characteristics and past adjustment results. By using the interactive genetic algorithm for the solution vector in that region, it becomes possible to eliminate the adjustment in the region that does not need to be adjusted, and to perform the optimal adjustment of the solution vector promptly. It has an excellent effect that

Further, the optimization adjusting method and the optimization adjusting apparatus of the second example related to the present invention, when using the interactive genetic algorithm, based on the past adjustment information recorded, the solution vector By setting the initial set of and eliminating and adjusting the solution vector that is not preferable to the user, the convergence to the optimal solution vector can be accelerated and the load when evaluating each solution vector can be reduced. Have the effect.

The optimization adjusting method and the optimization adjusting apparatus of the third example related to the present invention correct the evaluation value of the user based on the psychological condition estimated based on the physiological information of the user. This reduces the influence of fluctuations on the user's evaluation, so that the user's situation (hearing,
It has an excellent effect that the optimal solution vector adjustment for visual acuity etc. can be realized by an interactive genetic algorithm.

Further, the optimization adjusting method and the optimization adjusting apparatus of the fourth example related to the present invention are obtained when the user adjusts the optimum solution vector for each person by using the interactive genetic algorithm. The evaluation model for the user's adjustment process is estimated based on the adjustment history, and it has an excellent effect that the user's preference for this problem can be known.

Further, the optimization adjusting method and the optimization adjusting device of the fifth example related to the present invention are based on the history of solution vector optimization performed by a plurality of users using an interactive genetic algorithm. , Which has an excellent effect of being able to extract a common factor of tastes by a plurality of users.

The optimization adjusting method and the optimization adjusting apparatus of the sixth example related to the present invention use an interactive genetic algorithm to quickly execute the optimum solution vector adjustment for the user, and Since it works to update the common model regarding the adjustment process of the user, there is an excellent effect that it is possible for the user to reduce the burden on the user without having to evaluate the information by all the solution vectors.

The optimization adjusting method and the optimization adjusting apparatus of the seventh example related to the present invention estimate the evaluation model concerning the adjustment process of the user by using the interactive genetic algorithm, and obtain the obtained evaluation model. Since it automatically adjusts the optimal solution vector using, it is possible to reduce the burden on the user, which has been a problem when using the interactive genetic algorithm. Have.

Further, the optimization adjusting method and the optimization adjusting apparatus of the eighth example related to the present invention present information and information to assist in comparing / evaluating each solution vector in the problem of dealing with time series signals. The optimal solution vector is adjusted by adding the function of rearranging the order to the interactive genetic algorithm, which has been applied only to static data such as images until now. It has an excellent effect that the algorithm can be applied to dynamic data such as time series data.

Further, the optimization adjusting method and the optimization adjusting apparatus of the ninth example related to the present invention present information and information to assist in comparing / evaluating each solution vector in the problem of dealing with time series signals. It has the function of rearranging the order in which the adjustment is performed, and by using the interactive genetic algorithm by limiting the update region of each adjustment solution vector based on its characteristics and past adjustment results, The interactive genetic algorithm applied only to static data has an effect that it becomes possible to quickly adjust the optimum solution vector when targeting dynamic data such as time series data.

Further, the optimization adjusting method and the optimization adjusting apparatus of the tenth example related to the present invention present information and information for assisting in comparison / evaluation of solution vectors in the problem of dealing with time series signals. Has the function of rearranging the order in which
By using the interactive genetic algorithm by setting the initial set of solution vectors based on the recorded past adjustment information, the interactive genetic algorithm, which was previously applied only to static data such as images, is used. This has the effect of making it possible to quickly adjust the optimal solution vector when the dynamic algorithm targets dynamic data such as time-series data.

The optimization adjusting method and the optimization adjusting apparatus of the eleventh example related to the present invention present information and information to assist in comparing / evaluating each solution vector in the problem of dealing with time series signals. Has the function of rearranging the order in which
By updating the evaluation value of the user based on the psychological condition estimated based on the physiological information of the user, the optimal solution vector is updated by the interactive genetic algorithm while reducing the influence of fluctuation in the evaluation of the user. By
It becomes possible to apply the interactive genetic algorithm, which was previously applied only to static data such as images, to dynamic data such as time series data, and at the same time, the user's own This has the effect of making it possible to adjust the optimum solution vector corresponding to the situation.

The optimization adjusting method and the optimization adjusting apparatus of the twelfth example related to the present invention present information that assists in comparison and evaluation of solution vectors in the problem of handling time series signals. By rearranging the order of presentation according to the evaluation value in an interactive genetic algorithm, the evaluation model for the user's adjustment process is estimated based on the user's adjustment history. This has an effect of being able to know the state of preference for the time-series signal problem.

The optimization adjusting method and the optimization adjusting apparatus of the thirteenth example related to the present invention present information to assist in comparing / evaluating each solution vector in the problem of dealing with time series signals. And, by rearranging the order of presentation according to the evaluation value for an interactive genetic algorithm, a common model for the adjustment process of multiple users is estimated based on the recorded adjustment history of multiple users. Therefore, there is an effect that a common factor of preference when handling time-series signals by a plurality of users can be extracted.

The optimization adjusting method and the optimization adjusting apparatus of the fourteenth example related to the present invention present information that assists in comparison and evaluation of solution vectors in the problem of handling time series signals. And the function of rearranging the order of presentation according to the evaluation value, promptly execute the optimal solution vector adjustment for the user by using the interactive genetic algorithm, and the common model for the adjustment process of multiple users. Since it works so as to update, there is an effect that the user does not need to evaluate the information based on all the solution vectors and the burden on the user can be reduced.

Further, the optimization adjusting method and the optimization adjusting apparatus of the fifteenth example related to the present invention present information to assist in comparing / evaluating each solution vector in the problem of dealing with time series signals. And, it has a function to rearrange the order of presentation according to the evaluation value. The interactive genetic algorithm is used to estimate the evaluation model of the user's adjustment process, and the obtained individual model is used to automatically optimize the optimal solution. Since the vector is adjusted, it has the effect of reducing the burden on the user, which has been a problem when using the interactive genetic algorithm, and enabling application to time-series signals.

[0236]

As described above, the first optimization adjusting method and the optimization adjusting apparatus of the present invention, for each solution vector in the solution vector set, re-initializes the solution vector in the vicinity of the vector. A vector group is set, and a recombination operation is performed a fixed number of times to optimize the solution vector group in a limited area. Then, the plurality of obtained solution vector groups are integrated into one large set, and the solution vector is optimized again by the recombination operation. Therefore, after locally updating the solution vector by the recombination operation, the solution vector is updated again. By updating the global solution vector by the recombination operation, it is possible to reinforce the local update ability of the solution vector and perform a high-speed optimal solution estimation process.

Further, the second optimization adjusting method and the optimization adjusting device of the present invention compare each solution vector in the solution vector set with a vector group randomly extracted from its neighboring space and have a high degree of conformity. The group is selected again as a solution vector belonging to the solution vector set to be updated.
Then, by having the resulting solution vector set as the target of the recombination operation, there is an excellent effect that the weakness of the local updating ability, which is lacking in the conventional genetic algorithm, can be eliminated.

Further, the third optimization adjusting method and the optimization adjusting device of the present invention extract the neighborhood vector group of the solution vector selected by the goodness of fit, reset the solution set, and obtain the obtained neighborhood vector. By performing the recombination operation of the solution vector for the group, it is possible to enhance the local update capability and to prevent the influence of the solution vector having a high degree of conformity from having a great influence on the entire solution immediately.

Further, the fourth optimization adjusting method and the optimization adjusting device of the present invention divide the neighborhood of each solution vector in the solution vector set into a plurality of areas and randomly select a plurality of areas from the area having the largest average goodness of fit. A solution vector group is selected, and a new solution vector set is generated by a recombination operation on the solution vectors in the original solution vector set. Then, by selecting from these in descending order of goodness of fit and resetting one solution vector set, the solution vectors in the neighborhood of each solution vector are parallel to the optimization of the solution vector by the recombination operation in the solution vector set. It is said that the update process of the algorithm is also performed at the same time, and the local solution update capability can be reinforced to the efficient global solution update capability originally possessed by the genetic algorithm, and the optimal solution vector can be estimated efficiently. Has excellent effect.

Further, the fifth optimization adjusting method and the optimization adjusting device of the present invention include a solution set dividing section for dividing the solution vector set into a plurality of groups based on the arithmetic mean and the standard deviation of the goodness of fit and the solution. By providing the group recombination operation unit that performs the recombination operation of the solution vector for each group obtained by the set division unit, the influence of the solution vector with a high degree of conformity has a great influence on the entire solution immediately. It is possible to avoid this and to strengthen the local update ability.

Further, the sixth optimization adjusting method and the optimization adjusting device of the present invention are the arithmetic mean and standard deviation of the goodness of fit, the maximum
Based on the minimum goodness of fit, it is determined whether or not the solution vector set is divided into a plurality of groups. If it is determined to be divided, the solution set division unit divides the entire solution set into a plurality of groups, and if not so, the entire solution set is the target of the recombination operation, and the solution for the obtained recombination processing target is solved. By selecting a vector and performing a recombination operation, it is determined whether there is a solution vector with a high degree of goodness of conformity based on the goodness of fit distribution situation, and by limiting the recombination processing target so as to reduce the effect, efficiency can be improved. A good optimal solution can be estimated.

Further, the seventh optimization adjusting method and the optimization adjusting device of the present invention dynamically change the step convergence criterion for judging the convergence of the solution vector, and update the new solution vector. Has an excellent effect that an efficient optimal solution estimation can be realized by dynamically changing the value of Eq.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an optimization adjustment processing device according to a reference example of the present invention.

FIG. 2 is a block diagram showing a configuration of a recombination operation unit that is a main part of an optimization adjustment apparatus according to a reference example of the present invention.

FIG. 3 is a flowchart showing the overall processing steps of the optimization adjusting method in the reference example of the present invention.

FIG. 4 is a flowchart showing a process in process 1 of the optimization adjusting method in the reference example of the present invention.

FIG. 5 is a flowchart showing a process of a recombination operation process, which is a main process of the optimization adjusting method in the reference example of the present invention.

FIG. 6 is a conceptual diagram of a multidimensional function maximum value estimation problem treated as a specific example.

FIG. 7 is a diagram showing how a value in a real value section is converted into a fixed length bit string code.

FIG. 8 is a diagram schematically showing a solution vector update along an update direction vector.

FIG. 9 is a diagram for explaining a roulette selection method used for selective selection.

FIG. 10 is a block diagram showing the configuration of an optimization adjustment device according to the first embodiment of the present invention.

FIG. 11 is a flowchart showing the processing steps of the optimization adjusting method according to the first embodiment of the present invention.

FIG. 12 is a diagram schematically showing local update processing and global update processing.

FIG. 13 is a block diagram showing a configuration of an optimization adjustment device according to a second exemplary embodiment of the present invention.

FIG. 14 is a flowchart showing the processing steps of the optimization adjusting method according to the second embodiment of the present invention.

FIG. 15 is a diagram schematically showing how the initial solution vector group is extracted.

FIG. 16 is a block diagram showing a configuration of an optimization adjustment device according to a third exemplary embodiment of the present invention.

FIG. 17 is a flowchart showing the processing steps of the optimization adjusting apparatus according to the third embodiment of the present invention.

FIG. 18 is a diagram schematically showing a neighborhood vector extraction process in a real-valued space.

FIG. 19 is a block diagram showing a configuration of an optimization adjustment device according to a fourth exemplary embodiment of the present invention.

FIG. 20 is a flowchart showing the processing steps of the optimization adjusting method according to the fourth embodiment of the present invention.

FIG. 21 is a diagram schematically showing update area division and solution vector extraction processing from the division.

FIG. 22 is a block diagram showing the configuration of an optimization adjustment device according to a fifth embodiment of the present invention.

FIG. 23 is a flowchart showing the processing steps of the optimization adjusting method according to the fifth embodiment of the present invention.

FIG. 24 is a diagram schematically showing a solution set division process in a real-valued space. (A) The figure which represents typically the individual number distribution with respect to fitness fk. (B) A diagram showing a state of division in a real-valued space.

FIG. 25 is a block diagram showing the configuration of an optimization adjustment device according to a sixth embodiment of the present invention.

FIG. 26 is a flowchart showing the overall processing steps of the optimization adjusting method according to the sixth embodiment of the present invention.

FIG. 27 is a flowchart showing a process of process 2 of the optimization adjusting method according to the sixth embodiment of the present invention.

FIG. 28 is a block diagram showing the configuration of an optimization adjusting device according to a seventh embodiment of the present invention.

FIG. 29 is a flowchart showing the processing steps of the optimization adjusting method according to the seventh embodiment of the present invention.

FIG. 30 is a diagram schematically showing localization processing of an update area in a real value space.

FIG. 31 is a conceptual diagram of a one-dimensional function approximation problem.

FIG. 32 is a block diagram showing a configuration of an optimization adjustment device in a first example related to the present invention.

FIG. 33 is a flowchart showing the processing steps of the optimization adjusting method in the first example related to the present invention.

FIG. 34 is a block diagram showing the configuration of an optimization adjusting apparatus in a second example related to the present invention.

FIG. 35 is a flowchart showing the processing steps of the optimization adjusting method in the second example related to the present invention.

FIG. 36 is a block diagram showing the configuration of an optimization adjusting apparatus in a third example related to the present invention.

FIG. 37 is a flowchart showing the processing steps of the optimization adjusting method in the third example related to the present invention.

FIG. 38 is a block diagram showing the configuration of an optimization adjusting apparatus in a fourth example related to the present invention.

FIG. 39 is a flowchart showing the processing steps of the optimization adjusting method in the fourth example related to the present invention.

FIG. 40 is a block diagram showing the configuration of an optimization adjusting apparatus in a fifth example related to the present invention.

FIG. 41 is a flowchart showing the processing steps of the optimization adjusting method in the fifth example related to the present invention.

FIG. 42 is a flowchart showing the steps of an individual adjustment process of the optimization adjustment method in the fifth example related to the present invention.

FIG. 43 is a block diagram showing the configuration of an optimization adjusting apparatus in a sixth example related to the present invention.

FIG. 44 is a flowchart showing the processing steps of the optimization adjusting method in the sixth example related to the present invention.

FIG. 45 is a flowchart showing the steps of a common model update process of the optimization adjusting method in the sixth example related to the present invention.

FIG. 46 is a block diagram showing the configuration of an optimization adjusting apparatus in a seventh example related to the present invention.

FIG. 47 is a flowchart showing the processing steps of the optimization adjusting method in the seventh example related to the present invention.

FIG. 48 is a flowchart showing the continuation of the processing steps of the optimization adjusting method in the seventh example related to the present invention.

FIG. 49 is a block diagram showing the configuration of an optimization adjusting apparatus in an eighth example related to the present invention.

FIG. 50 is a flowchart showing the processing steps of the optimization adjusting method in the eighth example related to the present invention.

FIG. 51 is a block diagram showing the configuration of an optimization adjusting apparatus in a ninth example related to the present invention.

FIG. 52 is a flowchart showing the processing steps of the optimization adjusting method in the ninth example related to the present invention.

FIG. 53 is a block diagram showing the configuration of an optimization adjustment apparatus in a tenth example related to the present invention.

FIG. 54 is a flowchart showing the processing steps of the optimization adjusting method in the tenth example related to the invention.

FIG. 55 is a block diagram showing the configuration of an optimization adjusting apparatus in an eleventh example related to the present invention.

FIG. 56 is a flowchart showing the processing steps of the optimization adjusting method in the eleventh example related to the present invention.

FIG. 57 is a block diagram showing the configuration of an optimization adjustment apparatus in a twelfth example related to the present invention.

FIG. 58 is a flowchart showing the processing steps of the optimization adjusting method in the twelfth example related to the present invention.

FIG. 59 is a block diagram showing the configuration of an optimization adjusting apparatus in a thirteenth example related to the present invention.

FIG. 60 is a flowchart showing the processing steps of the optimization adjusting method in the thirteenth example related to the present invention.

FIG. 61 is a flowchart showing the process of individual adjustment processing 2 of the optimization adjustment method in the thirteenth example related to the present invention.

FIG. 62 is a block diagram showing the configuration of an optimization adjustment adjusting device in a fourteenth example related to the present invention.

FIG. 63 is a flowchart showing the processing steps of the optimization adjusting method in the fourteenth example related to the present invention.

FIG. 64 is a flowchart showing the steps of a common model updating process 2 of the optimization adjusting method in the fourteenth example related to the present invention.

FIG. 65 is a block diagram showing the configuration of an optimization adjusting apparatus in a fifteenth example related to the present invention.

FIG. 66 is a flowchart showing the processing steps of the optimization adjusting method in the fifteenth example related to the present invention.

FIG. 67 is a block diagram showing the configuration of a conventional optimization adjusting device using a genetic algorithm.

FIG. 68 is a flowchart showing the processing steps of a conventional optimization adjustment method using a genetic algorithm.

FIG. 69 is a block diagram showing the configuration of an optimization adjusting apparatus using a conventional interactive genetic algorithm.

FIG. 70 is a flowchart showing the processing steps of a conventional optimization adjustment method using an interactive genetic algorithm.

FIG. 71 is an explanatory diagram of the concept of the recombination operation in the genetic algorithm. (A) is a conceptual diagram showing an example of crossover processing.
(B) is a conceptual diagram showing an example of mutation processing.

FIG. 72 is a conceptual diagram of an adjustment problem of a vision correction lens.

FIG. 73 is a conceptual diagram of updating area limitation in a solution vector space.

FIG. 74 is a conceptual diagram of simplex crossover.

FIG. 75 is a diagram showing an example of a psychological situation estimation method of the optimization adjusting device in the third and eleventh examples related to the present invention. (A) Relationship between blinking frequency, attention, and interest (b) Relationship between sweating status, tension, and agitation

FIG. 76 is a configuration diagram of a neural network used in the evaluation model estimation unit in the optimization adjustment device in the fourth, seventh, 12, 15th examples related to the present invention.

[Fig. 77] Fig. 77 is a configuration diagram of a neural network used in a common model estimation unit in the optimization adjustment devices in the fifth, sixth, thirteenth, and fourteenth examples related to the present invention.

FIG. 78 is a conceptual diagram of a filter creation problem for improving the sound quality of distorted speech.

FIG. 79 is a conceptual diagram of auxiliary information corresponding to a time-series signal.

FIG. 80 is a conceptual diagram of a characteristic adjustment problem of a hearing aid for a hearing-impaired person.

[Explanation of symbols]

101 Initial Solution Set Setting Section 102 Genetic Algorithm Processing Unit 103 Optimal solution output section 104 Evaluation value acquisition unit 105 Fitness calculation section 106 update direction determination unit 107 Direction application update unit 108 Recombination operation unit 109 Center of gravity estimation unit 110 Update direction candidate recording unit 111 Update direction acquisition unit 112 Center of gravity moving part 113 Center of Gravity Solution Vector Generation Unit 201 Candidate selection part 202 Crossover processing execution unit 203 Mutation processing execution unit 204 selection range deriving unit 205 random number generator 206 Solution Vector Extractor 1001 Local update unit 1002 Global Update Department 1003 Local update setting unit 1004 Vector group initial setting section 1005 Local recombination operation unit 1006 Local update end determination unit 1007 Set Integration Department 1008 Global recombination operation unit 1301 initial update area limitation unit 1302 initial solution vector group extraction unit 1303 Solution vector set integration unit 1601 Solution set resetting unit 1602 Group recombination operation unit 1603 Representative solution vector selection unit 1604 Neighborhood vector extraction unit 1901 update area division unit 1902 average goodness-of-fit calculation unit 1903 Appropriate region solution vector extraction unit 1904 Solution vector integration unit 2201 Solution set division unit 2202 division area determination unit 2203 division execution unit 2501 Control unit to be rearranged 2801 update area setting unit 2802-step convergence determination unit 2803 Convergence criterion changing unit 3201 Target data input section 3202 update area limitation unit 3203 Main processing unit 3204 Equipment adjustment execution unit 3205 User Evaluation Department 3206 Set resetting section 3207 Information presentation section 3208 user evaluation determination unit 3401 Storage medium section 3402 Recorded information reading unit 3403 Initial solution vector selection unit 3404 initial solution vector supplement 3405 Optimal solution vector recording unit 3601 user psychology estimation unit 3602 evaluation value correction unit 3603 physiological data measurement unit 3604 Psychological estimation execution unit 3801 Evaluation model output unit 3802 Second User Evaluation Department 3803 Evaluation model estimation determination unit 3804 Model estimation execution unit 3805 Adjustment history recording unit 4001 Common model output unit 4002 user adjustment end determination unit 4003 Common model estimation unit 4004 Common model estimation determination unit 4005 Common model estimation execution unit 4301 Common model evaluation calculation unit 4302 Common model evaluation determination unit 4303 Common model update unit 4304 Common model update determination unit 4305 Common model update execution unit 4601 Method selection switch 4602 Evaluation model estimation unit 4603 model evaluation calculation unit 4604 Method switching determination unit 4901 Time Series Evaluation Execution Unit 4902 Auxiliary information presentation section 4903 Information sorting section 6701 selection selection execution unit 6901 Interactive Genetic Algorithm Execution Unit

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Kazuaki Ohara             1006 Kadoma, Kadoma-shi, Osaka Matsushita Electric             Sangyo Co., Ltd. (72) Inventor Susumu Maruno             1006 Kadoma, Kadoma-shi, Osaka Matsushita Electric             Sangyo Co., Ltd. F-term (reference) 5B056 BB62 BB64 BB91

Claims (22)

[Claims] 1. An optimization adjustment method for estimating an optimum solution by sequentially updating a solution vector arbitrarily set for a certain problem according to a goodness of fit by a predetermined arithmetic device, The first step to set the initial set, and set the initial solution vector group around each vector in the solution vector set, and rearrange the solution vectors based on the goodness of fit of the solution vector belonging to the initial solution vector group A second step of repeatedly performing the operation a preset number of times, and a genetic recombination targeting a solution vector in a solution vector set obtained by integrating a plurality of solution vector groups obtained in the second step into one. It has a third step of performing a predetermined recombination operation based on a procedure, and a fourth step of outputting an optimum solution vector, and satisfies a termination condition given in advance. In the second step, the local update of the solution vector is repeatedly performed, and then the global solution vector is updated in the third step to promptly estimate the optimum solution vector. Adjustment method. 2. An optimization adjusting method for estimating an optimum solution by sequentially updating a solution vector arbitrarily set for a certain problem according to a goodness of fit by a predetermined arithmetic device, A first step of setting an initial set, a second step of extracting a plurality of solution vectors from around each vector in the solution vector set, and a third step of obtaining an evaluation value for the solution vector group extracted in the second step , The fourth step of finding the goodness of fit of each solution vector from each evaluation value, and extracting the solution vector having a high goodness of fit from the extracted solution vector group and the solution vector at the center of the extracted area A fifth step of resetting the set and a predetermined vector based on the genetic recombination procedure based on the goodness of fit for the solution vector set reset in the fifth step. It has a sixth step of performing a recombination operation of, and a seventh step of outputting an optimal solution vector. By repeatedly executing the second step to the sixth step until the end condition given in advance is satisfied, Each solution vector in the solution vector set of is compared with the surrounding solution vector and replaced with a more optimal solution vector, and the solution vector is updated by recombination operation based on the completed solution vector set to estimate the optimal solution vector. An optimization adjusting method characterized by: 3. An optimization adjustment method for estimating an optimum solution by successively updating a solution vector arbitrarily set for a certain problem by a predetermined arithmetic unit according to the goodness of fit. A first step of setting an initial set; a second step of obtaining an evaluation value of a solution vector in the solution vector set for the problem; a third step of obtaining a goodness of fit of each solution vector from each evaluation value; Based on the 4th step of resetting the solution set by selecting the solution vector from the solution vector set and extracting the neighborhood vector group of the solution vector, and the vector in the neighborhood vector group obtained by the 4th step. The fifth step of performing a predetermined recombination operation based on the genetic recombination procedure based on each goodness of fit, and the solution to the new solution vector set obtained in the fifth step It has a sixth step of performing a mutation operation of randomly replacing a part of the vector with another symbol, and a seventh step of outputting an optimal solution vector, which is given in advance from the second step to the seventh step. The recombination operation is performed on the solution vectors in the plurality of neighborhood vector groups obtained in the fourth step until the end condition is satisfied, and then the mutation operation in the sixth step is performed on the entire solution vector set. An optimization adjusting method characterized by estimating an optimum solution vector by executing the method. 4. An optimization adjusting method for estimating an optimum solution by sequentially updating a solution vector arbitrarily set for a certain problem according to a goodness of fit by a predetermined arithmetic device, A first step of setting an initial set; a second step of obtaining an evaluation value of a solution vector in the solution vector set for the problem; a third step of calculating a goodness of fit of each solution vector from each evaluation value; The fourth step of performing an arithmetic recombination operation based on the genetic operation of the solution vector in the solution vector set according to the degree, and dividing the neighborhood space of multiple solution vectors arbitrarily selected from the solution vector set into multiple regions The fifth step, the sixth step of extracting a plurality of solution vector points in each area divided in the fifth step, and averaging the evaluation values and the goodness of fit thereof; A seventh step of selecting a region having the highest average goodness of fit and arbitrarily selecting a plurality of solution vectors from the region, one of the solution vector group selected in the seventh step and the solution vector group acquired in the fourth step It has an eighth step of setting a solution vector set, and a ninth step of outputting an optimal solution vector. The second to eighth steps are repeated until the preset end condition is satisfied, and the solution vector An optimization adjustment method characterized in that not only updating by recombining solution vectors in a set but also updating solution vectors in the vicinity of each solution vector in parallel is performed to estimate an optimal solution vector. . 5. An optimization adjusting method for estimating an optimum solution by sequentially updating a solution vector arbitrarily set for a certain problem according to a goodness of fit by a predetermined arithmetic device, A first step of setting an initial set; a second step of obtaining an evaluation value of the solution vector in the solution vector set for the problem; a third step of calculating a goodness of fit of each solution vector from each evaluation value; A fourth step of dividing the solution vector set into a plurality of groups based on the arithmetic mean and standard deviation of each goodness of fit obtained in the step, and a solution vector for the solution vector in the group obtained in the fourth step 5th step of generating a new solution vector set by performing an arithmetic recombination operation based on the genetic recombination of, and the group obtained in the 5th step is taken as one new solution. It has a sixth step of performing a mutation operation in which a part of the solution vector is randomly replaced with another symbol by integrating it into a vector set, and a seventh step of outputting an optimal solution vector. The second to sixth steps are repeated until the condition is satisfied, and the recombination operation is performed on the solution vectors in each group divided in the third step, and then the entire solution vector set is targeted in the sixth step. An optimal adjustment method characterized by estimating an optimum solution vector by executing the mutation operation described above. 6. An optimization adjusting method for estimating an optimum solution by sequentially updating a solution vector arbitrarily set for a certain problem according to a goodness of fit by a predetermined arithmetic device, A first step of setting an initial set; a second step of obtaining an evaluation value of a solution vector in the solution vector set for the problem; a third step of calculating a goodness of fit of each solution vector from each evaluation value; A fourth step of determining a recombination operation target based on genetic recombination based on the arithmetic mean and standard deviation of vector goodness of fit, maximum goodness of fit, and minimum goodness of fit, and in the fourth step, to multiple groups of the entire solution vector set When the division of is decided, it is decided in the 5th step which divides the solution vector set based on the arithmetic mean and standard deviation of the evaluation values, and in the 4th step based on each goodness of fit. It has a sixth step of generating a new solution vector set for the vector in the operation target, and a seventh step of outputting an optimal solution vector. From the second step until the preset end condition is satisfied. The sixth step is repeated, and whether or not to divide the solution vector set is controlled in the fourth step, and the recombination operation is performed on the operation target obtained as a result, so that the optimum solution vector is estimated. Optimization optimization method. 7. An optimization adjustment method for estimating an optimum solution by sequentially updating a solution vector arbitrarily set for a certain problem according to a goodness of fit by a predetermined arithmetic device, and taking a solution vector A first step of limiting the update region; a second step of setting an initial set of solution vectors in the update region obtained in the first step; and a step of obtaining an evaluation value of the solution vector in the solution vector set for the problem. 3 steps, 4th step of calculating the goodness of fit of each solution vector from each evaluation value, and arithmetic recombination operation based on genetic recombination of solution vectors in the solution vector set based on the goodness of fit of each solution vector. The fifth step of generating a new solution vector set by performing the above, and only the solution vectors included in the set update region in the group obtained in the fifth step 6th step of newly creating a solution vector set using, and dynamically changing the step convergence criterion for determining the convergence of the solution vector in the update vector set in the update region, and generating in the 5th step It comprises a seventh step of dynamically changing the updated area of the new solution vector, and an eighth step of outputting the optimum solution vector. From the third step to the seventh step until the preset end condition is satisfied. The steps are repeated, and the update area of the solution vector is first roughly set globally by the seventh step to update the solution vector, and then the update area is subdivided step by step to shift to the local update of the solution vector. The optimization adjustment method is characterized in that the optimal solution vector is estimated by performing the following. 8. An initial solution set setting section for setting an initial set of solution vectors to be updated, an initial solution vector group is set from around each vector in the solution vector set, and the solution vectors belonging to the initial solution vector group are set. A local update unit that repeatedly executes the recombining operation of the solution vector based on the goodness of fit to the problem, and a solution obtained by integrating a plurality of solution vector groups obtained by the local update unit. It is equipped with a global update unit that performs recombination operations of solution vectors in the vector set, and an optimal solution output unit that outputs an optimal solution obtained from the latest improved solution vector set when a preset termination condition is satisfied. Therefore, the local updating unit repeatedly updates the solution vector locally, and then the global updating unit updates the global solution vector to estimate an optimal solution. Optimization adjusting device sequentially repeated physical until satisfying the end condition. 9. A local update setting unit that determines a range in which the local update unit updates a local solution vector from a given solution vector, and a vector group that newly generates a solution vector group in the local update setting unit. An initial setting unit, an evaluation value acquisition unit that obtains an evaluation value for the target problem of the solution vector,
A goodness-of-fit calculation unit that calculates a goodness-of-fit of each solution vector from the evaluation value acquired by the evaluation-value acquisition unit, and a re-combination operation of solution vectors in a solution vector set based on the goodness-of-fit by the local update setting unit. It is configured by a local recombination operation unit that is executed so as to be included in the set range, and a local update end determination unit that determines whether or not the series of processing up to the above satisfies a predetermined number of iterations. 13. The optimization adjustment device according to claim 12, wherein the optimization adjustment device is a device. 10. A global updating section integrates each solution vector group obtained in the local updating section into one set, and recombination of solution vectors in the solution vector set obtained by the set integrating section. The optimization adjusting apparatus according to claim 8, wherein the optimization adjusting apparatus is configured by a global recombination operation unit that performs an operation. 11. The local recombination operation unit performs a genetic operation process represented by crossover, mutation, and selection used in a genetic algorithm.
The optimization adjusting device described in. 12. The global recombination operation unit performs a genetic operation process represented by crossover, mutation, and selection used in a genetic algorithm.
0. The optimization adjusting device described in 0. 13. An initial solution set setting unit that sets an initial set of solution vectors to be updated, an initial update region limiting unit that limits the periphery of each vector in the solution vector set, and an initial update limiting unit that is included in the initial update limiting unit. An initial solution vector group extraction unit that extracts a plurality of solution vectors, an evaluation value acquisition unit that acquires an evaluation value for the solution vector group obtained by the initial solution vector group extraction unit, and an evaluation value obtained by the evaluation value acquisition unit A goodness-of-fit calculation unit that calculates the goodness-of-fit of each solution vector from the values, and a solution vector with a high goodness of fit is extracted from the solution vectors that are the center of the extracted solution vector group and its extraction area, and the solution vector is newly created. A solution vector set integration unit that is a set, and a solution vector set created by the solution vector set integration unit is used as a target based on the goodness of fit obtained by the goodness of fit calculation unit. It has a recombination operation section that executes recombination operations and an optimal solution output section that outputs the optimal solution obtained from the updated latest solution vector set when the preset termination condition is satisfied. The process of comparing each solution vector in the set with the surrounding solution vector and replacing it with a more optimal solution vector, and updating the solution vector based on the completed solution vector set, and estimating the optimal solution vector An optimization adjusting apparatus that sequentially repeats the steps until the end condition is satisfied. 14. An initial solution set setting unit that sets an initial set of solution vectors to be updated, an evaluation value acquisition unit that obtains an evaluation value for the target problem of the solution vector, and an evaluation value acquired by the evaluation value acquisition unit. A goodness-of-fit calculation unit that calculates the goodness of fit of each solution vector, and a solution set that resets the solution set by selecting a solution vector from the solution vector set based on the goodness of fit and extracting a neighborhood vector group of the solution vector A resetting unit, and a group recombination operation unit that performs a manipulation operation of solution vectors based on the goodness of fit obtained by the goodness-of-fit calculation unit for the vectors in the neighborhood vector group obtained by the solution set resetting unit. And a mutation processing operation for performing a recombination operation of replacing a part of the solution vector with another symbol for the new solution vector set obtained by the group recombination operation unit. Section, and an optimal solution output section that outputs an optimal solution obtained from the latest solution vector set that has been improved when a preset termination condition is satisfied, and a plurality of solution sets obtained by the solution set resetting section. After performing the recombination operation on the solution vector in the neighborhood vector group of, the process of executing the recombination operation on the entire solution vector set in the mutation processing execution unit is repeated until the end condition is satisfied Optimization Adjustment device. 15. The solution set resetting unit defines a neighborhood based on the number of inversion bits when the solution vector is converted into a fixed-length bit string code, and extracts the neighborhood vector. Optimal adjustment device. 16. An initial solution set setting unit that sets an initial set of solution vectors to be updated, an evaluation value acquisition unit that obtains an evaluation value for the target problem of the solution vector, and an evaluation value acquired by the evaluation value acquisition unit. A goodness-of-fit calculation unit that calculates the goodness-of-fit of each solution vector, a shuffle operation unit that performs a shuffle operation of the solution vector in the solution vector set by the goodness-of-fit, and a plurality of solution vectors arbitrarily selected from the solution vector set. An update region dividing unit that divides the neighborhood space into a plurality of regions, and a plurality of solution vector points in each region divided by the update region dividing unit to extract the evaluation value and the average of the goodness of fit calculation of the goodness of fit Part, a region having the highest average goodness of fit among all regions, and an appropriate region solution vector extracting unit that arbitrarily selects a plurality of solution vectors from the region, and the solution vector extracting unit. Solution vector group and a solution vector integration section that sets one solution vector set from the solution vector group acquired by the recombination operation section, and the latest solution vector updated when a preset termination condition is satisfied. It is equipped with an optimal solution output section that outputs the optimal solution obtained from the set, and not only does the recombination operation section improve by recombining the solution vectors in the solution vector set, but also in the vicinity of each solution vector in parallel An optimization adjusting apparatus that sequentially repeats a process of updating a solution vector until the end condition is satisfied. 17. The updating area dividing unit divides into three areas, an area near the origin, a line segment connecting the two solution vectors, and an area farthest from the origin in a straight line section connecting the two selected solution vectors. It is divided into line segment areas, and the size of each area is 2
The optimization adjusting device according to claim 16, wherein the optimization adjusting device corresponds to a distance between two solution vectors. 18. An initial solution set setting unit that sets an initial set of solution vectors to be updated, an evaluation value acquisition unit that obtains an evaluation value for the target problem of the solution vector, and an evaluation value acquired by the evaluation value acquisition unit. A goodness-of-fit calculation unit that calculates the goodness of fit of each solution vector, a solution set division unit that divides the solution vector set into a plurality of groups based on the arithmetic mean and standard deviation of the goodness of fit, and the goodness-of-fit calculation unit A group recombination operation unit that generates a new solution vector set by performing a recombination operation of solution vectors on the solution vectors in the group obtained by the solution set division unit based on the obtained goodness of fit, and the group recombination A mutation processing execution unit that performs a recombination operation of replacing a part of the solution vector with another symbol for the entire new solution vector set obtained by the operation unit, and preset It is equipped with an optimal solution output unit that outputs an optimal solution obtained from the latest solution vector set that has been improved when the satisfied termination condition is satisfied, and targets the solution vector in the group obtained by the solution set division unit. The optimization adjusting apparatus that repeats the process of executing the recombination operation for the entire solution vector set in the mutation process execution unit after executing the recombination operation as described above until the end condition is satisfied. 19. An initial solution set setting unit that sets an initial set of solution vectors to be updated, an evaluation value acquisition unit that obtains an evaluation value for the target problem of the solution vector, and an evaluation value acquired by the evaluation value acquisition unit. A goodness-of-fit calculation unit that calculates the goodness-of-fit of each solution vector; a reshuffling target control unit that determines a reshuffling operation target based on the arithmetic mean and standard deviation of the goodness-of-fit, the maximum goodness-of-fit, and the minimum goodness of fit; A solution set division unit that divides the solution vector set based on the arithmetic mean and standard deviation of the evaluation values when the target control unit determines the division of the entire solution vector set into multiple groups, and the fitness calculation unit. The recombination operation unit that generates a new solution vector set for the vector in the operation target obtained by the recombination target control unit based on the goodness of fit obtained in It is equipped with an optimal solution output unit that outputs an optimal solution obtained from the improved latest solution vector set when satisfied, and controls whether or not to divide the solution vector set by the recombination target control unit. An optimization adjusting apparatus that repeats a process of executing a rearrangement operation on the obtained operation target until the end condition is satisfied. 20. An initial solution set setting unit that sets an initial set of solution vectors to be updated, an evaluation value acquisition unit that obtains an evaluation value for the target problem of the solution vector, and an evaluation value acquired by the evaluation value acquisition unit. A goodness of fit calculation unit that calculates the goodness of fit of each solution vector, and a new solution vector set is generated by performing a recombination operation of solution vectors in the solution vector set based on the goodness of fit obtained by the goodness of fit calculation unit. A recombination operation unit to perform, and an update area setting unit that dynamically changes the step convergence criterion for determining the convergence of the solution vector and dynamically changes the update area of the new solution vector generated by the recombination operation unit. And an optimum solution output unit that outputs an optimum solution obtained from the latest solution vector set improved when a preset termination condition is satisfied, First, the update region setting unit globally rearranges the update vector in each element constituting the solution vector, and rearranges the solution vector, and then gradually tightens the step convergence criterion in the update region and subdivides the region. An optimization adjusting apparatus that repeats a process of locally updating a solution vector by performing optimization until the end condition is satisfied. 21. The optimization according to claim 14, wherein the group recombination operation unit uses a genetic algorithm for performing a genetic operation process represented by crossover, mutation, and selection. Adjustment device. 22. The recombination operation unit includes crossover, mutation,
A genetic algorithm for performing a predetermined arithmetic processing based on a gene level recombination operation represented by selection and selection is used.
The optimization adjusting device according to any one of 1.