WO2025054840A1 - Information transmission method and apparatus - Google Patents

Abstract

Embodiments of the present application provide an information transmission method and apparatus. The method comprises: a first device determines first adjustment information on the basis of a first neural network, the first adjustment information being used for indicating adjustment information of at least one neural network layer in a second neural network, and the second neural network being a neural network in a second device. The first device sends the first adjustment information to the second device. The second device adjusts the at least one neural network layer in the second neural network on the basis of the first adjustment information. Hence, the first device can use the first neural network to assist the second device to adjust the second neural network, so that the second neural network adjusted by the second device can adapt to a communication scenario, thereby facilitating improvement of the generalization ability of the second neural network. For a scenario in which the second neural network is used by the second device to process transmitted data, the method is conducive to improving the generalization ability of the neural network used for processing the transmitted data in a communication system.

Description

Information transmission method and device Technical Field

The present application relates to the field of communication technology, and in particular to an information transmission method and device.

Background Art

Machine learning is a technical means to achieve artificial intelligence. Neural networks are a type of machine learning. Neural networks have the ability to infinitely approximate any continuous function given by the universal approximation theorem, and can accurately abstractly model complex high-dimensional problems.

Neural networks can be applied to both the sending and receiving ends of a communication system to process transmitted data. The sending end can use a neural network to process the data to be sent and send the processed data to the receiving end; the receiving end can use a neural network to process the data received from the sending end. How to improve the generalization ability of the neural network used to process the transmitted data in the communication system is a technical problem that needs to be solved.

Summary of the invention

The embodiments of the present application provide an information transmission method and device, which are beneficial to improving the generalization capability of a neural network used to process transmitted data in a communication system.

In a first aspect, the present application provides an information transmission method, which can be applied to a first device, a chip in the first device, or a logic module or software that can implement all or part of the functions of the first device. The first device is used as an example for description below. The method includes: the first device determines first adjustment information based on a first neural network, and the first adjustment information is used to indicate adjustment information of at least one neural network layer in a second neural network, and the second neural network is a neural network in a second device. The first device sends the first adjustment information to the second device, and the first adjustment information is used by the second device to adjust at least one neural network layer in the second neural network.

It can be seen that the first device can use the first neural network to assist the second device in adjusting the second neural network. Based on the first neural network, the first adjustment information adapted to the current communication scenario can be determined, which is conducive to the second neural network adjusted by the second device to adapt to the current communication scenario. The first neural network can be used to determine the first adjustment information for different scenarios, which is conducive to the second device being able to adaptively adjust the second neural network for different scenarios, improving the universality of the second neural network, so that the second neural network has the ability to generalize in different scenarios. This information transmission method can be applied to the scenario where the second neural network is used for the second device to process the transmitted data, which is conducive to improving the generalization ability of the neural network used to process the transmitted data in the communication system.

In addition, compared to the data expansion method of improving the generalization ability of the second neural network by expanding the training data set of the second neural network, the information transmission method provided in the embodiment of the present application uses the first neural network to assist the second device in adjusting the second neural network, which can reduce the performance degradation of the second neural network caused by insufficient feature capture due to the complexity and diversity of wireless scenarios in the data expansion method. In addition, the information transmission method provided in the embodiment of the present application enables efficient online adjustment of the neural network layer in the second neural network without retraining the second neural network offline, which can simultaneously ensure network performance and real-time requirements.

In an optional implementation, the first adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structure adjustment information of at least one neural network layer in the second neural network. It can be seen that the first device can assist the second device in adjusting the weight and/or bias and/or activation function and/or structure of the second neural network by using the first neural network.

In an optional implementation, the first adjustment information is obtained by inputting information used to characterize the distribution of channel characteristics into the output of the first neural network. Alternatively, the first adjustment information is obtained by inputting information used to characterize the distribution of channel characteristics and structural information of at least one neural network layer in the second neural network into the output of the first neural network. It can be seen that the input of the first neural network includes information used to characterize the distribution of channel characteristics, which is conducive to making the second neural network adjusted by the second device based on the first adjustment information more adaptable to the channel, thereby improving the communication quality.

It can be understood that in this information transmission method, for multiple scenarios with different channel feature distributions, the first neural network can be used to determine the first adjustment information for adjusting the second neural network in each scenario based on the channel feature distribution in each scenario, which is conducive to improving the universality of the second neural network, so that the second neural network has generalization characteristics in scenarios with different channel feature distributions, so that the adjusted second neural network can quickly adapt to the distribution of its input information and improve the adaptability of data and scenarios.

Optionally, the information used to characterize the channel characteristic distribution includes one or more of the following: channel state information, noise distribution information, signal-to-noise ratio, delay distribution information, power distribution information, Doppler spread information, and reference signal received power (RSRP).

In an optional implementation, before the first device determines the first adjustment information based on the first neural network, the method further includes: the first device receives first indication information from the second device, the first indication information is used to apply for adjusting the second neural network, or the first indication information Used to request activation of the first neural network. It can be seen that the second device can trigger the adjustment of the second neural network or trigger the activation of the first neural network.

In another optional implementation, before the first device determines the first adjustment information based on the first neural network, the method further includes: the first device sends first indication information to the second device, the first indication information is used to apply for adjustment of the second neural network, or the first indication information is used to request activation of the first neural network. It can be seen that the first device can trigger the adjustment of the second neural network or trigger the activation of the first neural network.

In an optional implementation, the method further includes: the first device sending second indication information to the second device, the second indication information being used to indicate at least one neural network layer to be adjusted in the second neural network. It can be seen that the first device can negotiate with the second device to determine the neural network layer to be adjusted in the second neural network.

In another optional implementation, the method further includes: the first device receiving second indication information from the second device, the second indication information being used to indicate at least one neural network layer to be adjusted in the second neural network. It can be seen that the first device can negotiate with the second device to determine the neural network layer to be adjusted in the second neural network.

In an optional embodiment, the method further includes: the first device determines second adjustment information based on the first neural network, and the second adjustment information is used to indicate adjustment information of at least one neural network layer in the third neural network. The first device adjusts at least one neural network layer in the third neural network based on the second adjustment information. It can be seen that the first device can also use the first neural network to assist in adjusting the third neural network included in itself, which is beneficial to improving the universality of the third neural network, so that the third neural network has the ability to generalize in different scenarios. This information transmission method can be applied to the scenario where the third neural network is used by the first device to process the transmitted data, which is beneficial to improving the generalization ability of the neural network used to process the transmitted data in the communication system.

Optionally, the second adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structure adjustment information of at least one neural network layer in the third neural network.

In the second aspect, the present application provides an information transmission method, which can be applied to a second device, a chip in the second device, or a logic module or software that can realize all or part of the functions of the second device. The second device is used as an example for description below. The method includes: the second device receives first adjustment information from the first device, the first adjustment information is used to indicate adjustment information of at least one neural network layer in the second neural network, and the first adjustment information is determined by the first device based on the first neural network. The second device adjusts at least one neural network layer in the second neural network based on the first adjustment information.

It can be seen that the first device can use the first neural network to assist the second device in adjusting the second neural network. Based on the first neural network, the first adjustment information adapted to the current communication scenario can be determined, which is conducive to the second neural network adjusted by the second device to adapt to the current communication scenario. The first neural network can be used to determine the first adjustment information for different scenarios, which is conducive to the second device being able to adaptively adjust the second neural network for different scenarios, improving the universality of the second neural network, and making the second neural network have the ability to generalize in different scenarios. This information transmission method can be applied to the scenario where the second neural network is used for the second device to process the transmitted data, which is conducive to improving the generalization ability of the neural network used to process the transmitted data in the communication system.

In addition, compared to the data expansion method of improving the generalization ability of the second neural network by expanding the training data set of the second neural network, the second device in the information transmission method provided in the embodiment of the present application uses the first adjustment information output by the first neural network to adjust the second neural network, which can reduce the performance degradation of the second neural network caused by insufficient feature capture due to the complexity and diversity of the wireless scene in the data expansion method. In addition, the information transmission method provided in the embodiment of the present application enables efficient online adjustment of the neural network layer in the second neural network without retraining the second neural network offline, which can simultaneously ensure network performance and real-time requirements.

In an optional implementation, the first adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structure adjustment information of at least one neural network layer in the second neural network. It can be seen that the first device can use the first neural network to assist the second device in adjusting the weight and/or bias and/or activation function and/or structure of the second neural network.

In an optional implementation, the first adjustment information is obtained by the first device inputting information used to characterize the distribution of channel characteristics into the first neural network. Alternatively, the first adjustment information is obtained by the first device inputting information used to characterize the distribution of channel characteristics and structural information of at least one neural network layer in the second neural network into the first neural network. It can be seen that the input of the first neural network includes information used to characterize the distribution of channel characteristics, which is conducive to making the second neural network adjusted by the second device based on the first adjustment information more adaptable to the channel, thereby improving communication quality.

It can be understood that in this information transmission method, for multiple scenarios with different channel feature distributions, the first neural network can be used to determine the first adjustment information for adjusting the second neural network in each scenario based on the channel feature distribution in each scenario, which is conducive to improving the universality of the second neural network, so that the second neural network has generalization characteristics in scenarios with different channel feature distributions, so that the adjusted second neural network can quickly adapt to the distribution of its input information and improve the adaptability of data and scenarios.

Optionally, the information used to characterize the channel characteristic distribution includes one or more of the following: channel state information, noise distribution information, signal-to-noise ratio, delay distribution information, power distribution information, Doppler spread information, and RSRP.

In an optional implementation manner, before the second device receives the first adjustment information from the first device, the method further includes: the second device The first device sends a first indication message to the first device, where the first indication message is used to apply for adjusting the second neural network, or the first indication message is used to request activation of the first neural network. It can be seen that the second device can trigger the adjustment of the second neural network or trigger the activation of the first neural network.

In another optional implementation, before the second device receives the first adjustment information from the first device, the method further includes: the second device receives the first indication information from the first device, the first indication information is used to apply for adjusting the second neural network, or the first indication information is used to request activation of the first neural network. It can be seen that the first device can trigger the adjustment of the second neural network or trigger the activation of the first neural network.

In an optional implementation, the method further includes: the second device receiving second indication information from the first device, the second indication information being used to indicate at least one neural network layer to be adjusted in the second neural network. It can be seen that the second device and the first device can negotiate to determine the neural network layer to be adjusted in the second neural network.

In another optional implementation, the method further includes: the second device sending second indication information to the first device, the second indication information being used to indicate at least one neural network layer to be adjusted in the second neural network. It can be seen that the second device can negotiate with the first device to determine the neural network layer to be adjusted in the second neural network.

In a third aspect, the present application further provides a communication device. The communication device has the function of implementing some or all of the implementation methods described in the first aspect above, or has the function of implementing some or all of the functional implementation methods described in the second aspect above. The functions can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more units or modules corresponding to the above functions.

In one possible design, the structure of the communication device may include a processing unit and a communication unit, wherein the processing unit is configured to support the communication device to perform corresponding functions in the above method. The communication unit is used to support communication between the communication device and other communication devices. The communication device may also include a storage unit, which is used to couple with the processing unit and the communication unit and store necessary program instructions and data for the communication device.

In one implementation, the communication device includes: a processing unit and a communication unit, and the processing unit is used to control the communication unit to send and receive data/signaling.

A processing unit is used to determine first adjustment information based on the first neural network, where the first adjustment information is used to indicate adjustment information of at least one neural network layer in a second neural network, and the second neural network is a neural network in a second device.

A communication unit is used to send first adjustment information to a second device, where the first adjustment information is used by the second device to adjust at least one neural network layer in a second neural network.

In addition, in this aspect, other optional implementations of the communication device can refer to the relevant content of the first aspect mentioned above and will not be described in detail here.

In another implementation, the communication device includes: a processing unit and a communication unit, and the processing unit is used to control the communication unit to send and receive data/signaling.

A communication unit is used to receive first adjustment information from a first device, where the first adjustment information is used to indicate adjustment information of at least one neural network layer in a second neural network, and the first adjustment information is determined by the first device based on the first neural network.

A processing unit is used to adjust at least one neural network layer in the second neural network based on the first adjustment information.

In addition, in this aspect, other optional implementations of the communication device can refer to the relevant content of the second aspect mentioned above and will not be described in detail here.

As an example, the communication unit may be a transceiver or a communication interface, the storage unit may be a memory, and the processing unit may be a processor.

In one embodiment, the communication device includes: a processor and a transceiver. The processor is used to determine first adjustment information based on a first neural network, the first adjustment information is used to indicate adjustment information of at least one neural network layer in a second neural network, and the second neural network is a neural network in a second device. The transceiver is used to send the first adjustment information to the second device, and the first adjustment information is used by the second device to adjust at least one neural network layer in the second neural network.

In addition, in this aspect, other optional implementations of the communication device can refer to the relevant content of the first aspect mentioned above and will not be described in detail here.

In another embodiment, the communication device includes: a processor and a transceiver. The transceiver is used to receive first adjustment information from the first device, the first adjustment information is used to indicate adjustment information of at least one neural network layer in the second neural network, and the first adjustment information is determined by the first device based on the first neural network. The processor is used to adjust at least one neural network layer in the second neural network based on the first adjustment information.

In addition, in this aspect, other optional implementations of the communication device can refer to the relevant content of the first aspect mentioned above and will not be described in detail here.

In another embodiment, the communication device is a chip or a chip system. The processing unit may also be embodied as a processing circuit or a logic circuit; the transceiver unit may be an input/output interface, an interface circuit, an output circuit, an input circuit, a pin or a related circuit on the chip or the chip system.

In the implementation process, the processor can be used to perform, for example but not limited to, baseband related processing, and the transceiver or communication interface can be used to perform, for example but not limited to, radio frequency transceiver. The above devices can be respectively arranged on independent chips, or at least partially or completely arranged on the same chip. For example, the processor can be further divided into an analog baseband processor and a digital baseband processor. Among them, the analog baseband The processor can be integrated with the transceiver (or communication interface) on the same chip, and the digital baseband processor can be set on an independent chip. With the continuous development of integrated circuit technology, more and more devices can be integrated on the same chip. For example, a digital baseband processor can be integrated with a variety of application processors (such as but not limited to a graphics processor, a multimedia processor, etc.) on the same chip. Such a chip can be called a system on a chip (SoC). Whether each device is independently set on different chips or integrated on one or more chips often depends on the needs of product design. The embodiments of the present application do not limit the implementation form of the above-mentioned devices.

In a fourth aspect, the present application also provides a processor for executing the above-mentioned various methods. In the process of executing these methods, the process of sending the above-mentioned signal and receiving the above-mentioned signal in the above-mentioned method can be understood as the process of outputting the above-mentioned signal by the processor, and the process of the above-mentioned signal input by the processor. When outputting the above-mentioned signal, the processor outputs the above-mentioned signal to the transceiver so that it can be transmitted by the transceiver (or communication interface). After the above-mentioned signal is output by the processor, it may also need to perform other processing before it reaches the transceiver (or communication interface). Similarly, when the processor receives the above-mentioned input signal, the transceiver (or communication interface) receives the above-mentioned signal and inputs it into the processor. Furthermore, after the transceiver (or communication interface) receives the above-mentioned signal, the above-mentioned signal may need to perform other processing before it is input into the processor.

For the sending and receiving operations involved in the processor, unless otherwise specified, or unless they conflict with their actual function or internal logic in the relevant description, they can be more generally understood as processor output, reception, input and other operations, rather than sending and receiving operations performed directly by the RF circuit and antenna.

In the implementation process, the processor may be a processor specifically used to execute these methods, or a processor that executes computer instructions in a memory to execute these methods, such as a general-purpose processor. The memory may be a non-transitory memory, such as a read-only memory (ROM), which may be integrated with the processor on the same chip or may be separately arranged on different chips. The embodiment of the present application does not limit the type of memory and the arrangement of the memory and the processor.

In a fifth aspect, the present application further provides a communication system, which includes at least one first device and at least one second device of the above aspects. In another possible design, the system may also include other devices that interact with the first device and/or the second device in the solution provided by the present application.

In a sixth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is run, the method described in the first aspect or the second aspect is executed.

In a seventh aspect, the present application further provides a computer program product comprising instructions, wherein the computer program product comprises: a computer program code, and when the computer program code is run, the method described in the first aspect or the second aspect above is executed.

In an eighth aspect, the present application provides a chip system, which includes a processor and an interface, wherein the interface is used to obtain a program or instruction, and the processor is used to call the program or instruction to implement the function involved in the first aspect, or to call the program or instruction to implement the function involved in the second aspect. In one possible design, the chip system also includes a memory, which is used to store program instructions and data necessary for the terminal. The chip system can be composed of a chip, or it can include a chip and other discrete devices.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of a communication system provided in an embodiment of the present application;

FIG2 is a schematic diagram of a feedforward neural network provided in an embodiment of the present application;

FIG3 is a schematic diagram of a recursive neural network provided in an embodiment of the present application;

FIG4 is a schematic diagram of a hidden layer of a convolutional neural network provided in an embodiment of the present application;

FIG5 is a schematic diagram of an end-to-end network signal processing provided by an embodiment of the present application;

FIG6 is a schematic diagram of a physical layer architecture provided in an embodiment of the present application;

FIG7 is a schematic diagram of another physical layer architecture provided in an embodiment of the present application;

FIG8 is a schematic diagram of a flow chart of an information transmission method provided in an embodiment of the present application;

FIG9 is a schematic diagram of an adaptive intelligent architecture network provided in an embodiment of the present application;

FIG10 is a schematic diagram of another adaptive intelligent architecture network provided in an embodiment of the present application;

FIG11 is a schematic diagram of another information transmission method provided in an embodiment of the present application;

FIG12 is a schematic diagram of another information transmission method provided in an embodiment of the present application;

FIG13 is a schematic diagram of another information transmission method provided in an embodiment of the present application;

FIG14 is a schematic diagram of another information transmission method provided in an embodiment of the present application;

FIG15 is a schematic diagram of simulation results of a test using Gaussian noise provided in an embodiment of the present application;

FIG16 is a schematic diagram of simulation results of a test using mixed Gaussian noise provided in an embodiment of the present application;

FIG17 is a schematic diagram of the structure of a communication device provided in an embodiment of the present application;

FIG18 is a schematic diagram of the structure of another communication device provided in an embodiment of the present application.

DETAILED DESCRIPTION

The technical solutions in the embodiments of the present application are clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

In order to better understand the information transmission method disclosed in the embodiment of the present application, a communication system applicable to the embodiment of the present application is described.

The technical solution of the embodiment of the present application can be applied to various communication systems. For example, the global mobile communication system, the long term evolution (LTE) system, the universal mobile communication system, the fourth generation mobile communication technology (4th generation, 4G) system, the next generation radio access network (NG-RAN), the new radio (NR) system, the fifth generation mobile communication technology (5th generation mobile networks, 5G) system, and with the continuous development of communication technology, the technical solution of the embodiment of the present application can also be used for subsequent evolution of communication systems, such as the sixth generation mobile communication technology (6th generation mobile networks, 6G) system, the seventh generation mobile communication technology (7th generation mobile networks, 7G) system, and so on. The technical solution of the embodiment of the present application is applicable to the new air interface transmission scenario of the future communication system and supports adaptive intelligent architecture network.

Please refer to Figure 1, which is a schematic diagram of the structure of a communication system provided in an embodiment of the present application, and the communication system includes but is not limited to a first device and a second device. The number and form of the devices shown in Figure 1 are used for example and do not constitute a limitation on the embodiment of the present application. In actual applications, two or more first devices and two or more second devices may be included. Among them, the first device may be a network device or a terminal device, and the second device may be a network device or a terminal device.

In the embodiment of the present application, the network device has a wireless transceiver function, and the network device includes but is not limited to: a base station (BS), a radio network controller (RNC), a network device controller (BSC), a network device transceiver station (BTS), a home network device (e.g., home evolved Node B, or home Node B, HNB), a baseband unit (BBU), a wireless relay node, a wireless backhaul node, a transmission point (TRP; or, TP), a satellite, etc. Among them, a base station is a device deployed in a wireless access network that can provide wireless communication functions, which can also be called a base station device, for example, an evolutionary base station (evolutional Node B, eNB or e-NodeB) in a long term evolution (LTE) system, a node B (Node B), a base station (gNodeB or gNB) in a 5G system, a base station in a 6G system, etc. A base station can include a BBU and a remote radio unit (RRU). The BBU and RRU can be placed in different places, for example: the RRU is remote and placed in an area with high traffic volume, and the BBU is placed in a central computer room. The BBU and RRU can also be placed in the same computer room. The BBU and RRU can also be different components under the same rack. Base stations can be in the following forms: macro base stations, micro base stations (also called small stations), pico base stations, relay stations, access points, balloon stations, etc.

The terminal device may also be referred to as user equipment (UE), terminal, access terminal, subscriber unit, user station, mobile station, mobile station (MS), remote station, remote terminal, mobile device, user terminal, user agent or user device, and may be applied to 4G, 5G or even 6G systems. The terminal device in the embodiment of the present application may be a handheld device, vehicle-mounted device, wearable device, computing device or other processing device connected to a wireless modem with wireless communication function; the terminal device may be a terminal with the function of connecting to a cellular base station. For example, the terminal device may be a cellular phone, a smart phone, a tablet computer, a wireless data card, a personal digital assistant (PDA) computer, a tablet computer, a wireless modem, a handheld device (handset), a laptop computer, a machine type communication (MTC) terminal, etc. The terminal device can also be a virtual reality (VR) terminal device, an augmented reality (AR) terminal device, a wireless terminal in industrial control, a wireless terminal in self driving, a wireless terminal in remote medical, a wireless terminal in smart grid, a wireless terminal in transportation safety, a wireless terminal in a smart city, a wireless terminal in a smart home, a vehicle-mounted terminal, a wireless communication device in a smart factory, and so on.

The following is a brief introduction to the relevant concepts involved in the embodiments of the present application.

1. Explainable artificial intelligence (XAI)

The research objectives of XAI include: understanding the decision-making process of artificial intelligence (AI), fully understanding AI models, and developing translatable and explainable AI models. XAI's research areas include: making deep neural network components transparent, learning semantic networks from deep neural networks, and generating explanations (generating explanations that people can understand). XAI's research dimensions include: rationality, understandability, and traceability. Among them, rationality can be expressed as: the reasoning and prediction capabilities of AI models need to be reasonable. Understandability can be expressed as: people can understand the model on which AI decisions are based to a certain extent or completely. Traceability can be expressed as: the ability to track the reasoning or prediction process based on the nature of the data and the logic of the mathematical algorithm.

2. Deep neural network (DNN)

Neural network algorithm is a technology of machine learning, and machine learning is a technical means to achieve artificial intelligence. Neural network algorithm has the ability to infinitely approximate any continuous function given by the universal approximation theorem, and it can accurately abstractly model complex high-dimensional problems. Deep neural networks and deep learning based on neural network algorithms can be applied to image processing, speech processing, natural language processing, disease diagnosis, etc.

A deep neural network is generally a multi-layer structure. Increasing the depth and width of a neural network can improve its expressiveness, thereby providing a more powerful information extraction and abstract modeling capability for complex systems. For example, a feedforward neural network (FNN) is a deep neural network. FIG2 is a schematic diagram of a feedforward neural network provided in an embodiment of the present application. In conjunction with FIG2, the neural network layers in the FNN include an input layer, a hidden layer, and an output layer. Among them, the input layer can be used to input information to be processed. The hidden layer can be used to extract information features; a DNN generally contains at least two hidden layers, and the multiple hidden layers can be used to extract information features to different degrees. FIG2 is an example of one hidden layer. The output layer can be used to map the required output information from the extracted feature information. The circular patterns in FIG2 represent neurons of each layer, wherein the white circular pattern represents neurons of the input layer, the gray circular pattern represents neurons of the hidden layer, and the black circular pattern represents neurons of the output layer. The connection mode between neurons in each layer and the activation function used determine the expression function of the neural network.

Taking into account different needs, DNN can have a variety of construction methods. For example, recurrent neural networks (RNN), convolutional neural networks (CNN) and fully connected neural networks respectively represent a construction method of DNN. Figure 3 is a schematic diagram of a recursive neural network provided in an embodiment of the present application. In conjunction with Figure 3, the difference between RNN and the FNN shown in Figure 2 is that the output of the neurons in the RNN can directly act on itself at the next moment. For example, the input of the i-th layer of neurons at the nth moment includes the output of the i-1th layer of neurons at the nth moment, as well as the output of the i-th layer of neurons at the n-1th moment, as shown by the semicircular arrow in Figure 3.

FIG4 is a schematic diagram of a hidden layer of a convolutional neural network provided in an embodiment of the present application. The large black boxes in FIG4 are channels or feature maps of each layer, and the matrices represented by small boxes of different background colors are convolution kernels corresponding to different channels. For CNN, not all upper and lower layer neurons can be directly connected, but are connected through "convolution kernels" as intermediaries. In addition, in CNN, the information to be processed still retains its original positional relationship after the convolution operation.

3. End to End Learning (E2E Learning)

End-to-end learning is characterized by directly optimizing the overall goal of the task without performing module-based or phase-based training during deep learning. End-to-end learning can be applied to scenarios where both the sender and receiver process information in a communication system. In this scenario, an end-to-end joint optimization approach can be used to design the overall system, which can greatly improve transmission performance. Combined with Figure 5, when processing information, the sender and receiver can use a joint optimization approach to construct a problem model, or can use some auxiliary information between each other (for example, the sender can forward transmit auxiliary information to the receiver, and the receiver can reversely transmit auxiliary information to the sender) to achieve the purpose of improving overall performance.

When a neural network (such as a deep neural network) is used for joint optimization of the transmitter and receiver, the neural networks at both ends need to be designed together to ensure the convergence of the neural network in training and the excellent performance in reasoning. Specifically, the neural networks at both ends can be constructed by jointly constructing a cost function, or the neural networks at both ends can be constructed using an autoencoder structure or various variations of the autoencoder structure, or other neural network structures can be combined to construct the final neural network at both ends, without limitation.

Exemplarily, a physical layer architecture is shown in FIG6. The transmitter processes the signal to be sent through multiple modules such as channel coding, interleaving, scrambling, pilot insertion, OFDM modulation, CP insertion, layer mapping, precoding, BF weighting, etc., and then adds AWGN to the processed signal and sends it to the receiver through the wireless channel. The receiver processes the received signal through multiple modules such as frequency offset estimation and compensation, channel estimation, MIMO detection, equalizer, whitening filter, deinterleaving, channel decoding, etc. It can be seen that the physical layer architecture shown in FIG6 is complex and bloated. The end-to-end intelligent physical layer architecture can be shown in FIG7. The transmitter can select a set of parameters from multiple sets of parameters as the parameters of the modulation network, input the bits to be sent into the modulation network, and the modulation network processes the input bits; the transmitter sends the transmission symbols output by the modulation network to the receiver through the channel. The receiver can use the channel estimation network and the signal detection network to process the received symbols. During the processing, the channel estimation network and the signal detection network can be alternately iterated. Among them, the modulation network, the channel estimation network and the signal detection network are all neural networks. It can be seen that processing the transmitted data based on end-to-end learning neural networks at both the sending and receiving ends can simplify the processing process and improve the overall performance.

The embodiments of the present application are described below in conjunction with the accompanying drawings.

Please refer to Figure 8, which is a flow chart of an information transmission method provided in an embodiment of the present application. The diagram takes the first device and the second device as the execution subject of the interaction diagram as an example to illustrate the corresponding method, but the present application does not limit the execution subject of the interaction diagram. For example, the first device in the figure can also be a chip or chip system or processor that supports the first device to implement the corresponding method, or a logic module or software that can implement all or part of the functions of the first device. The second device in the figure can also be a chip or chip system or processor that supports the second device to implement the corresponding method, or a logic module or software that can implement all or part of the functions of the second device. The information transmission method includes the following steps.

S101. A first device determines first adjustment information based on a first neural network, where the first adjustment information is used to indicate adjustment information of at least one neural network layer in a second neural network.

Among them, at least one neural network layer in the second neural network is a partial neural network layer or all neural network layers in the second neural network. It is understandable that the first adjustment information can be used to indicate the adjustment information of a partial neural network layer or all neural network layers in the second neural network. In addition, the second neural network is a neural network in the second device, and the communication function carried by the second neural network is different from the communication function carried by the first neural network (the communication function carried by the first neural network includes the function for determining the first adjustment information). The embodiment of the present application does not limit the specific function of the second neural network. For example, the second neural network is used for the second device to process the data/signal to be sent. In this case, the second neural network can also be called a modulation network. For another example, the second neural network is used for the second device to process the received data/signal. In this case, the second neural network can also be called a detection network. For another example, the second neural network is used for the second device to compress channel state information. For another example, the second neural network is used for the second device to perform channel estimation. In addition, in an embodiment of the present application, the first neural network can also be called a dynamic adjustment network.

In an optional implementation, the first adjustment information is obtained by the first device inputting information used to characterize the distribution of channel characteristics into the first neural network. Alternatively, the first adjustment information is obtained by the first device inputting information used to characterize the distribution of channel characteristics and structural information of at least one neural network layer in the second neural network into the first neural network. It can be seen that the input of the first neural network includes information used to characterize the distribution of channel characteristics, which is conducive to making the second neural network adjusted by the second device based on the first adjustment information more adaptable to the channel, thereby improving communication quality.

It can be understood that in this information transmission method, for multiple scenarios with different channel feature distributions, the first neural network can be used to determine the first adjustment information for adjusting the second neural network in each scenario based on the channel feature distribution in each scenario, which is conducive to improving the universality of the second neural network, so that the second neural network has generalization characteristics in scenarios with different channel feature distributions, so that the adjusted second neural network can quickly adapt to the distribution of its input information and improve the adaptability of data and scenarios.

Optionally, the information used to characterize the distribution of channel characteristics includes one or more of the following: channel state information (CSI), noise distribution information, signal-to-noise ratio, delay distribution information, power distribution information, Doppler spread information, reference signal receiving power (RSRP). Optionally, the channel state information may be obtained by the first device performing channel estimation based on a reference signal from the second device. For example, in a scenario where the first device is a terminal device and the second device is a network device, the first device may perform channel estimation based on a channel state information reference signal (CSI reference signal, CSI-RS) from the second device to obtain the channel state information. For another example, in a scenario where the first device is a network device and the second device is a terminal device, the first device may perform channel estimation based on a demodulation reference signal (SRS) from the second device to obtain the channel state information.

In addition, the embodiments of the present application do not limit the content included in the information used to characterize the distribution of channel characteristics. Exemplarily, the information used to characterize the distribution of channel characteristics may include input information of the second neural network in addition to the aforementioned content. Alternatively, the information used to characterize the distribution of channel characteristics may also include signals of other modes, for example, it may include relevant information of the previous several moments for characterizing time correlation, other frequency band information for characterizing frequency domain correlation, other port information for characterizing spatial correlation, and so on.

In an optional embodiment, the first adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structure adjustment information of at least one neural network layer in the second neural network. Among them, the weight adjustment information indicated by the first adjustment information is used to adjust the weight of the neural network layer in the second neural network, the bias adjustment information indicated by the first adjustment information is used to adjust the bias of the neural network layer in the second neural network, the activation function indicated by the first adjustment information is used to adjust the activation function of the neural network layer in the second neural network, and the structure adjustment information indicated by the first adjustment information is used to adjust the structure of the neural network layer in the second neural network. In addition, when the first adjustment information indicates multiple items of weight adjustment information, bias adjustment information, activation function, and structure adjustment information, the multiple items indicated by the first adjustment information can be used to adjust the same neural network layer in the second neural network, or can be used to adjust different neural network layers in the second neural network, without limitation.

For example, the second neural network includes neural network layer #1 to neural network layer #3. The first adjustment information is specifically used to indicate weight adjustment information #1 of neural network layer #1 and bias adjustment information #1 of neural network layer #1 in the second neural network; wherein weight adjustment information #1 is used to adjust the weight of neural network layer #1, and bias adjustment information #1 is used to adjust the bias of neural network layer #1.

For another example, the second neural network includes neural network layers #1 to #3. The first adjustment information is specifically used to indicate the weight adjustment information #1 of neural network layer #1, the bias adjustment information #1 of neural network layer #1, and the bias adjustment information #1 of neural network layer #2 in the second neural network; wherein the weight adjustment information #1 is used to adjust the weight of neural network layer #1, the bias adjustment information #1 is used to adjust the bias of neural network layer #1, and the bias adjustment information #2 is used to adjust the bias of neural network layer #2.

In an optional implementation, before the first device determines the first adjustment information based on the first neural network, the method may further include: transmitting first indication information between the first device and the second device, thereby triggering the adjustment of the second neural network or triggering the activation of the first neural network. The first indication information is used to apply for adjusting the second neural network, or the first indication information is used to request activation of the first neural network. Triggering the adjustment of the second neural network or triggering the activation of the first neural network can be manifested as: triggering the first device to perform model training to obtain the first neural network; The network may be specifically described in the following optional implementation 1.1 and implementation 1.2.

In implementation 1.1, the second device sends the first indication information to the first device; accordingly, the first device receives the first indication information from the second device, and the first device performs model training to obtain the first neural network. It can be seen that the second device can trigger the adjustment of the second neural network or trigger the activation of the first neural network.

Optionally, in implementation 1.1, after the first device receives the first indication information from the second device, the method may further include: the first device sends a third indication information to the second device; accordingly, the first device receives the third indication information from the second device. Wherein, the third indication information is used to indicate confirmation information for the first indication information. Specifically, if the first indication information is used to apply for adjustment of the second neural network, the third indication information can be used to indicate confirmation information for adjusting the second neural network; if the first indication information is used to request activation of the first neural network, the third indication information can be used to indicate confirmation information for activating the first neural network. In addition, the embodiment of the present application does not restrict the order between the operation of the first device sending the third indication information to the second device and the operation of the first device performing model training to obtain the first neural network. For example, after the first device receives the first indication information from the second device, it can first send the third indication information to the second device, and then perform the operation of performing model training to obtain the first neural network. For another example, in a scenario where the first device is a network device (e.g., a base station), after the first device receives the first indication information from the second device, it can first perform the operation of performing model training to obtain the first neural network, and then send the third indication information to the second device.

Optionally, in implementation 1.1, the operation of the second device sending the first indication information to the first device may be performed periodically, on-demand, non-periodically, or semi-permanently (also referred to as semi-persistently). In an optional manner, steps S101 to S103 may be performed each time the operation of the second device sending the first indication information to the first device is performed. Optionally, steps S101 to S103 may be performed periodically each time the operation of the second device sending the first indication information to the first device is performed, until the second device sends a termination indication to the first device, and stops repeatedly performing steps S101 to S103.

In implementation 1.2, the first device sends the first indication information to the second device; correspondingly, the second device receives the first indication information from the first device. The first device performs model training to obtain the first neural network. It can be seen that the first device can trigger the adjustment of the second neural network or trigger the activation of the first neural network.

Optionally, in implementation 1.2, after the second device receives the first indication information from the first device, the method may further include: the second device sends a third indication information to the first device; accordingly, the first device receives the third indication information from the second device. Wherein, the third indication information is used to indicate confirmation information for the first indication information. Specifically, if the first indication information is used to apply for adjustment of the second neural network, the third indication information can be used to indicate confirmation information for adjusting the second neural network; if the first indication information is used to request activation of the first neural network, the third indication information can be used to indicate confirmation information for activating the first neural network. In addition, the embodiment of the present application does not restrict the order between the operation of the first device receiving the third indication information from the second device and the operation of the first device performing model training to obtain the first neural network. For example, after the first device sends the first indication information to the second device, it can perform the operation of performing model training to obtain the first neural network when/after receiving the third indication information from the second device. For another example, in a scenario where the first device is a network device (e.g., a base station), after the first device receives the first indication information from the second device, it can first perform model training to obtain the first neural network, and then receive the third indication information from the second device.

Optionally, in implementation 1.2, the operation of the first device sending the first indication information to the second device may be performed periodically, on-demand, non-periodically, or semi-permanently (also referred to as semi-persistently). In an optional manner, steps S101 to S103 may be performed each time the operation of the first device sending the first indication information to the second device is performed. Optionally, steps S101 to S103 may be performed periodically each time the operation of the first device sending the first indication information to the second device is performed, until the first device sends a termination indication to the second device, and stops repeatedly performing steps S101 to S103.

In an optional embodiment, the method may further include: the first device sends second indication information to the second device, the second indication information is used to indicate at least one neural network layer to be adjusted in the second neural network; accordingly, the second device receives the second indication information from the first device. Through this embodiment, the first device can negotiate with the second device to determine the neural network layer to be adjusted in the second neural network. Exemplarily, this embodiment can be applied to the scenario described in Implementation 1.1 where the second device sends the first indication information to the first device. After receiving the first indication information from the second device, the first device can perform the operation of sending the second indication information to the second device. In addition, the embodiment of the present application does not limit the order of the operation of the first device sending the second indication information to the second device and the operation of performing model training with the first device to obtain the first neural network.

In another optional implementation, the method may further include: the second device sends second indication information to the first device, the second indication information is used to indicate at least one neural network layer to be adjusted in the second neural network; accordingly, the first device receives the second indication information from the second device. Through this implementation, the second device can negotiate with the first device to determine the neural network layer to be adjusted in the second neural network. Exemplarily, this implementation can be applied to the scenario described in implementation 1.2 where the first device sends the first indication information to the second device. After receiving the first indication information from the first device, the second device can execute the operation of sending the second indication information to the first device. In addition, The embodiment of the present application does not restrict the order of the operation of the first device receiving the second indication information from the second device and the operation of performing model training with the first device to obtain the first neural network.

Optionally, the aforementioned second indication information may include an identifier (e.g., an index) of at least one neural network layer to be adjusted in the second neural network. Alternatively, the second indication information may include other information capable of identifying at least one neural network layer to be adjusted in the second neural network, without limitation.

S102: The first device sends first adjustment information to the second device; correspondingly, the second device receives the first adjustment information from the first device.

S103. The second device adjusts at least one neural network layer in the second neural network based on the first adjustment information.

For example, the first adjustment information is specifically used to indicate the weight adjustment information #1 of the neural network layer #1, the bias adjustment information #1 of the neural network layer #1, and the bias adjustment information #2 of the neural network layer #2 in the second neural network. Then, the second device can adjust the weight of the neural network layer #1 based on the weight adjustment information #1, adjust the bias of the neural network layer #2 based on the bias adjustment information #1, and adjust the bias of the neural network layer #2 based on the bias adjustment information #2.

For another example, the first device shown in FIG9 is a terminal device and the second device is a network device. In conjunction with FIG9, the second neural network in the network device is used by the network device to process data to be sent to the terminal device, and the third neural network in the terminal device is used by the terminal device to process data received from the network device; the second neural network and the third neural network both include a convolution (Conv) layer, a rectified linear unit (ReLU) layer, and a linear (Linear) layer, and the third neural network also includes a Sigmoid function layer. The input of the first neural network in the terminal device may include part/all of the input information of the third neural network. The terminal device determines the first adjustment information based on the first neural network, and the first adjustment information is specifically used to indicate the weight adjustment information #1 and the bias adjustment information #1 of the linear layer in the second neural network; the terminal device sends the first adjustment information to the network device. Then, the network device can adjust the weight of the linear layer in the second neural network layer based on the weight adjustment information #1, and adjust the bias of the linear layer in the second neural network layer based on the bias adjustment information #1.

In addition, the embodiment of the present application does not limit the specific adjustment method of the second device adjusting at least one neural network layer in the second neural network based on the first adjustment information. It is understandable that the embodiment of the present application does not limit the mathematical operations used in the process of the second device adjusting at least one neural network layer in the second neural network based on the first adjustment information. For example, the mathematical operations used may include one or more of the following: addition, multiplication, convolution, and direct substitution. The following is explained by taking the example of the second device adjusting the weights of at least one neural network layer in the second neural network based on the first adjustment information. The specific adjustment method of the second device adjusting the bias and structure of at least one neural network layer in the second neural network based on the first adjustment information is similar and will not be repeated.

For example, the first adjustment information is specifically used to indicate the weight adjustment information #1 (weight adjustment information #1 includes weight #1) of the neural network layer #1 in the second neural network, and the weight of the neural network layer #1 in the second neural network before adjustment is weight #2. The second device can use the result obtained by multiplying weight #1 and weight #2 as the adjusted weight of the neural network layer #1. Alternatively, the second device can use the result obtained by adding weight #1 and weight #2 as the adjusted weight of the neural network layer #1. Alternatively, the second device can use the result obtained by convolving weight #1 with weight #2 as the adjusted weight of the neural network layer #1. Alternatively, the second device can also use the result obtained by performing one or more other mathematical operations on weight #1 and weight #2 as the adjusted weight of the neural network layer #1 in the second neural network. Alternatively, the second device can also directly use weight #1 as the adjusted weight of the neural network layer #1. Alternatively, the second device can also use the result obtained by performing one or more mathematical operations on weight #1 as the adjusted weight of the neural network layer #1. No limitation is made.

In an optional implementation, the number of the first neural networks may be one or more. In the case where the number of the first neural networks is one, the first device may determine a first adjustment information based on the first neural network, and the second device may adjust the second neural network based on the first adjustment information. In the case where the number of the first neural networks is multiple, the first device may determine a first adjustment information based on each of the multiple first neural networks, and the second device may adjust the second neural network based on the multiple first adjustment information.

For example, the first device determines first adjustment information #1 based on the first neural network #1, and the first adjustment information #1 is used to indicate weight adjustment information #1 of neural network layer #1 in the second neural network and bias adjustment information #1 of neural network layer #1; the first device also determines first adjustment information #2 based on the first neural network #2, and the first adjustment information #2 is used to indicate weight adjustment information #2 of neural network layer #1 in the second neural network and bias adjustment information #2 of neural network layer #1. The first device may send the first adjustment information #1 and the first adjustment information #2 to the second device. The second device may adjust the weight of neural network layer #1 based on weight adjustment information #1 and weight adjustment information #2, and adjust the bias of neural network layer #1 based on bias adjustment information #1 and bias adjustment information #2.

In an optional embodiment, the method may also include: the first device determines second adjustment information based on the first neural network, the second adjustment information is used to indicate adjustment information of at least one neural network layer in the third neural network; the first device adjusts at least one neural network layer in the third neural network based on the second adjustment information.

Wherein, at least one neural network layer in the third neural network is a partial neural network layer or all neural network layers in the third neural network. It is understandable that the second adjustment information can be used to indicate adjustment information of a partial neural network layer or all neural network layers in the third neural network. In addition, The third neural network is a neural network in the first device. The communication function carried by the third neural network is different from the communication function carried by the first neural network. The embodiments of the present application do not limit the specific functions of the third neural network. For example, the third neural network is used by the first device to process data/signals to be sent. In this case, the third neural network can also be called a modulation network. For another example, the third neural network is used by the first device to process received data/signals. In this case, the third neural network can also be called a detection network. For another example, the third neural network is used by the first device to compress channel state information. For another example, the third neural network is used by the first device to perform channel estimation.

Optionally, the third neural network may be a neural network adapted to the second neural network. For example, the second neural network is used by the second device to process data/signals to be sent to the first device, and the third neural network is used by the first device to process data/signals received by the second device after being processed by the second neural network. For another example, the third neural network is used by the first device to process data/signals to be sent to the second device, and the second neural network is used by the second device to process data/signals received by the first device after being processed by the third neural network.

Optionally, the second adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structural adjustment information of at least one neural network layer in the third neural network. Optionally, the second adjustment information is obtained by inputting information used to characterize the distribution of channel characteristics into the first neural network; or, the second adjustment information is obtained by inputting information used to characterize the distribution of channel characteristics and structural information of at least one neural network layer in the third neural network into the first neural network. For the scenario in which the first device simultaneously determines the first adjustment information and the second adjustment information based on the first neural network, the first device inputs information used to characterize the distribution of channel characteristics into the first neural network, or the first device inputs information used to characterize the distribution of channel characteristics and structural information of at least one neural network layer in the second neural network and/or structural information of at least one neural network layer in the third neural network into the first neural network, and the first neural network can output the first adjustment information and the second adjustment information.

In addition, the second adjustment information is similar to the aforementioned first adjustment information, and the first device adjusts at least one neural network layer in the third neural network based on the second adjustment information, which is similar to the aforementioned second device adjusting at least one neural network layer in the second neural network based on the first adjustment information. For specific descriptions, please refer to the aforementioned related descriptions and will not be repeated here. In addition, for the implementation mode in which the first device determines the second adjustment information based on the first neural network, the first indication information mentioned in the aforementioned implementation modes 1.1 and 1.2 can also be used to apply for adjustment of the third neural network.

In an optional manner, the method may also include: the first device sends fourth indication information to the second device, the fourth indication information is used to indicate at least one neural network layer to be adjusted in the third neural network; accordingly, the second device receives the fourth indication information from the first device. In another optional manner, the method may also include: the second device sends fourth indication information to the first device, the fourth indication information is used to indicate at least one neural network layer to be adjusted in the third neural network; accordingly, the first device receives the fourth indication information from the second device. It is understandable that the first device and the second device can negotiate to determine the neural network layer to be adjusted in the third neural network. Optionally, the fourth indication information may include an identifier (e.g., an index) of at least one neural network layer to be adjusted in the third neural network. Alternatively, the fourth indication information may include other information that can identify at least one neural network layer to be adjusted in the third neural network, without limitation.

It is understandable that in the embodiment of the present application, the first device can determine the first adjustment information based on the first neural network, but not the second adjustment information. Alternatively, the first device can simultaneously determine the first adjustment information and the second adjustment information based on the first neural network. Alternatively, the first device can determine the second adjustment information based on the first neural network, but not the first adjustment information. Among them, the first adjustment information is used for the second device to adjust the second neural network, and the second adjustment information is used for the first device to adjust the third neural network. For the specific description of the first adjustment information and the second adjustment information, please refer to the above-mentioned related description, which will not be repeated here. In addition, for the case where the first device determines multiple first neural networks, the multiple first neural networks are all used to determine the first adjustment information but not the second adjustment information, or the multiple first neural networks are all used to determine the second adjustment information but not the first adjustment information, or some or all of the multiple first neural networks are used to determine the first adjustment information and the second adjustment information. The following is an example of the first device performing model training to obtain the first neural network #1 and the first neural network #2.

For example, the first device determines the first adjustment information #1 based on the first neural network #1, and determines the first adjustment information #2 based on the first neural network #2. The first adjustment information #1 and the first adjustment information #2 are both used by the second device to adjust the second neural network.

For another example, the first device determines the second adjustment information #1 based on the first neural network #1, and determines the second adjustment information #2 based on the first neural network #2. The second adjustment information #1 and the second adjustment information #2 are both used by the first device to adjust the third neural network.

For another example, the first device determines the first adjustment information #1 based on the first neural network #1, and determines the second adjustment information #2 based on the first neural network #2. The first adjustment information #1 is used by the second device to adjust the second neural network, and the second adjustment information #2 is used by the first device to adjust the third neural network.

For another example, the first device determines the first adjustment information #1 and the second adjustment information #1 based on the first neural network #1, and determines the first adjustment information #2 based on the first neural network #2. The first adjustment information #1 and the first adjustment information #2 are both used by the second device to adjust the second neural network, and the second adjustment information #1 is used by the first device to adjust the third neural network.

In an optional implementation, the method may further include: the second device performs model training to obtain a fourth neural network, and the second device determines the third adjustment information and/or the fourth adjustment information based on the fourth neural network. The third adjustment information is used to indicate at least one of the second neural network. The third adjustment information is used by the second device to adjust at least one neural network layer in the second neural network. The fourth adjustment information is used to indicate the adjustment information of at least one neural network in the third neural network, and the fourth adjustment information is used by the first device to adjust at least one neural network layer in the third neural network. It can be seen that the second device can also train the fourth neural network by itself to determine the adjustment information that can be used to adjust other neural networks.

In the case where the second device determines the third adjustment information based on the fourth neural network, the second device may adjust at least one neural network layer in the second neural network based on the third adjustment information. Optionally, if the second device also receives the first adjustment information from the first device, the second device may adjust at least one neural network layer in the second neural network based on the third adjustment information and the first adjustment information.

In the case where the second device determines the fourth adjustment information based on the fourth neural network, the second device also sends the fourth adjustment information to the first device. Then, the first device can adjust at least one neural network layer in the third neural network based on the fourth adjustment information. Optionally, if the first device also determines the second adjustment information based on the first device, then the first device can adjust at least one neural network layer in the third neural network based on the fourth adjustment information and the second adjustment information.

In addition, the number of fourth neural networks can be one or more, and any one of the one or more fourth neural networks can be used to determine the third adjustment information and/or the fourth adjustment information. The fourth neural network is similar to the first neural network, and the difference between the two is that the first neural network is a neural network trained by the first device, and the fourth neural network is a neural network trained by the second device. For the specific description of the fourth neural network, please refer to the aforementioned related description of the first neural network, which will not be repeated here. In addition, in the embodiment of the present application, the fourth neural network can also be referred to as a dynamic adjustment network, the first neural network is a dynamic adjustment network in the first device, and the fourth neural network is a dynamic adjustment network in the second device.

Optionally, the third adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structural adjustment information of at least one neural network layer in the second neural network. The fourth adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structural adjustment information of at least one neural network layer in the third neural network. Optionally, the third adjustment information and/or the fourth adjustment information is obtained by the second device inputting information used to characterize the distribution of channel characteristics into the fourth neural network and outputting it. Alternatively, the third adjustment information is obtained by the second device inputting information used to characterize the distribution of channel characteristics and structural information of at least one neural network layer in the second neural network into the fourth neural network and outputting it. Alternatively, the fourth adjustment information is obtained by the second device inputting information used to characterize the distribution of channel characteristics and structural information of at least one neural network layer in the third neural network and outputting it into the fourth neural network and outputting it. The third adjustment information is similar to the aforementioned second adjustment information, and the fourth adjustment information is similar to the aforementioned first adjustment information. For details, please refer to the aforementioned related explanations and will not be repeated here.

For example, in combination with FIG. 10, the first device determines the first adjustment information #1 and the second adjustment information #1 based on the first neural network #1, the first adjustment information #1 is specifically used to indicate the weight adjustment information #1 and the bias adjustment information #1 of the neural network layer #1 in the second neural network, and the second adjustment information #1 is specifically used to indicate the weight adjustment information #2 and the bias adjustment information #2 of the neural network layer #2 in the third neural network. The first device sends the first adjustment information #1 to the second device. The second device determines the third adjustment information #1 based on the fourth neural network #1, and the third adjustment information #1 is specifically used to indicate the weight adjustment information #3 and the bias adjustment information #3 of the neural network layer #1 in the second neural network. Then, the second device can adjust the weight of the neural network layer #1 in the second neural network based on the weight adjustment information #1 and the weight adjustment information #3, and adjust the bias of the neural network layer #1 in the second neural network based on the bias adjustment information #1 and the bias adjustment information #3. The first device can adjust the weight of the neural network layer #2 in the third neural network based on the weight adjustment information #2, and adjust the bias of the neural network layer #2 in the third neural network based on the bias adjustment information #2.

In an optional implementation, after the second device adjusts the second neural network and/or the first device adjusts the third neural network, it may also be performed that: the second device trains the adjusted second neural network and/or the first device trains the adjusted third neural network.

In addition, the embodiments of the present application do not impose any restrictions on the signaling used to carry any information (such as the first adjustment information, the fourth adjustment information, the first indication information, the second indication information, the third indication information, and the fourth indication information) transmitted between the first device and the second device mentioned above. The embodiments of the present application do not impose any restrictions on the transmission channel, transmission period, data quantization, and index format for transmitting any information between the first device and the second device mentioned above.

In summary, in the information transmission method, the first device can determine first adjustment information based on the first neural network, the first adjustment information is used to indicate adjustment information of at least one neural network layer in the second neural network, and the second neural network is a neural network in the second device. The first device sends the first adjustment information to the second device. The second device adjusts at least one neural network layer in the second neural network based on the first adjustment information.

It can be seen that the first device can use the first neural network to assist the second device in adjusting the second neural network. Based on the first neural network, the first adjustment information adapted to the current communication scenario can be determined, which is conducive to the second neural network adjusted by the second device to adapt to the current communication scenario. The first neural network can be used to determine the first adjustment information for different scenarios, which is conducive to the second device being able to adaptively adjust the second neural network for different scenarios, improving the universality of the second neural network, so that the second neural network has the ability to generalize in different scenarios. This information transmission method can be applied to the scenario where the second neural network is used for the second device to process the transmitted data, which is conducive to improving the generalization ability of the neural network used to process the transmitted data in the communication system.

In addition, compared to the data expansion method of improving the generalization ability of the second neural network by expanding the training data set of the second neural network, the information transmission method provided in the embodiment of the present application uses the first neural network to assist the second device in adjusting the second neural network, which can reduce the performance degradation of the second neural network caused by insufficient feature capture due to the complexity and diversity of wireless scenarios in the data expansion method. In addition, the information transmission method provided in the embodiment of the present application enables efficient online adjustment of the neural network layer in the second neural network without retraining the second neural network offline, which can simultaneously ensure network performance and real-time requirements.

The information transmission method provided in the embodiment of the present application is exemplarily described below in conjunction with specific scenarios.

The following is an example in which the first device is a terminal device, the second device is a network device, the second neural network is used by the network device to process data to be sent to the terminal device, and the third neural network is used by the terminal device to process data received from the network device. The information transmission method provided in the embodiment of the present application can be as described in the following examples 1 and 2.

Example 1: The terminal device triggers the adjustment of the second neural network and the third neural network or triggers the activation of the first neural network. As shown in FIG11 , the exemplary information transmission method includes the following steps S201 to S211 .

S201, the terminal device sends first indication information to the network device, the first indication information is used to apply for adjusting the second neural network and the third neural network, or the first indication information is used to request activation of the first neural network. Correspondingly, the network device receives the first indication information from the terminal device.

S202. The network device sends to the terminal device: third indication information, and/or, second indication information and fourth indication information. The third indication information is used to indicate confirmation information for the first indication information, the second indication information is used to indicate at least one neural network layer to be adjusted in the second neural network, and the fourth indication information is used to indicate at least one neural network layer to be adjusted in the third neural network. Accordingly, the terminal device receives from the network device: the third indication information, and/or, the second indication information and the fourth indication information.

S203: The terminal device performs model training to obtain a first neural network.

S204. The terminal device determines first adjustment information and second adjustment information based on the first neural network, wherein the first adjustment information is used to indicate adjustment information of at least one neural network layer in the second neural network, and the second adjustment information is used to indicate adjustment information of at least one neural network layer in the third neural network.

S205. The terminal device sends first adjustment information to the network device; correspondingly, the network device receives the first adjustment information from the terminal device.

S206. The network device adjusts at least one neural network layer in the second neural network based on the first adjustment information.

S207. The terminal device adjusts at least one neural network layer in the third neural network based on the second adjustment information.

In an optional implementation, the above steps S202 to S207 may be performed periodically until step S208 is performed, and the repetitive execution of steps S202 to S206 may be stopped. In step S208, the terminal device sends a termination instruction to the network device; correspondingly, the network device receives the termination instruction from the terminal device.

S209: The network device processes the data to be sent to the terminal device based on the adjusted second neural network.

S210: The network device sends data processed by the adjusted second neural network to the terminal device. Correspondingly, the terminal device receives data processed by the adjusted second neural network from the network device.

S211. The terminal device processes data from the network device based on the adjusted third neural network.

The embodiment of the present application does not limit the order between step S205 and step S207. The specific description of steps S201-S211 can refer to the related description in the aforementioned information transmission method, which will not be repeated here.

Example 2: The network device triggers the adjustment of the second neural network and the third neural network or triggers the activation of the first neural network. The difference between the information transmission method in this case and the information transmission method described in Example 1 is that step S201 is replaced by step S301, step S202 is replaced by step S302, and step S208 is replaced by step S303, as shown in the dotted box in Figure 12.

S301: A network device sends first indication information to a terminal device. Correspondingly, the terminal device receives the first indication information from the network device.

S302: The terminal device sends the third indication information, and/or the second indication information and the fourth indication information to the network device. Correspondingly, the network device receives the third indication information, and/or the second indication information and the fourth indication information from the terminal device.

S303: The network device sends a termination instruction to the terminal device. Correspondingly, the terminal device receives the termination instruction from the network device.

The following is an example in which the first device is a network device, the second device is a terminal device, the third neural network is used by the network device to process data to be sent to the terminal device, and the second neural network is used by the terminal device to process data received from the network device. The information transmission method provided in the embodiment of the present application can be as described in the following examples 3 and 4.

Example 3: The network device triggers the adjustment of the second neural network and the third neural network or triggers the activation of the first neural network. As shown in FIG13 , the exemplary information transmission method includes the following steps S401 to S411 .

S401, the network device sends first instruction information to the terminal device, the first instruction information is used to apply for adjusting the second neural network and the third neural network The network, or the first indication information is used to request activation of the first neural network. Accordingly, the terminal device receives the first indication information from the network device.

S402. The terminal device sends to the network device: third indication information, and/or, second indication information and fourth indication information. The third indication information is used to indicate confirmation information for the first indication information, the second indication information is used to indicate at least one neural network layer to be adjusted in the second neural network, and the fourth indication information is used to indicate at least one neural network layer to be adjusted in the third neural network. Accordingly, the network device receives from the terminal device: the third indication information, and/or, the second indication information and the fourth indication information.

S403: The network device performs model training to obtain a first neural network.

S404: The network device determines first adjustment information and second adjustment information based on the first neural network, wherein the first adjustment information is used to indicate adjustment information of at least one neural network layer in the second neural network, and the second adjustment information is used to indicate adjustment information of at least one neural network layer in the third neural network.

S405. The network device sends first adjustment information to the terminal device; correspondingly, the terminal device receives the first adjustment information from the network device.

S406. The terminal device adjusts at least one neural network layer in the second neural network based on the first adjustment information.

S407. The network device adjusts at least one neural network layer in the third neural network based on the second adjustment information.

In an optional implementation, the above steps S402 to S407 may be performed periodically until step S408 is performed, and the repetitive execution of steps S402 to S407 may be stopped. In step S408, the network device sends a termination instruction to the terminal device; correspondingly, the terminal device receives the termination instruction from the network device.

S409: The network device processes the data to be sent to the terminal device based on the adjusted third neural network.

S410: The network device sends data processed by the adjusted third neural network to the terminal device. Correspondingly, the terminal device receives data processed by the adjusted third neural network from the network device.

S411. The terminal device processes data from the network device based on the adjusted second neural network.

The embodiment of the present application does not limit the order between step S405 and step S407. The specific description of steps S401-S411 can refer to the related description in the aforementioned information transmission method, which will not be repeated here.

Example 4: The terminal device triggers the adjustment of the second neural network and the third neural network or triggers the activation of the first neural network. The difference between the information transmission method in this case and the information transmission method described in Example 3 is that step S401 is replaced by step S501, step S402 is replaced by step S502, and step S408 is replaced by step S503; in addition, after the network device receives the first indication information from the terminal device in step S501, it can first execute step S403 and step S404, and then execute step S502. As shown in the dotted box in Figure 14.

S501: The terminal device sends first indication information to the network device. Correspondingly, the network device receives the first indication information from the terminal device.

S502: The network device sends the third indication information and/or the second indication information and the fourth indication information to the terminal device. Correspondingly, the terminal device receives the third indication information and/or the second indication information and the fourth indication information from the network device.

S503: The terminal device sends a termination instruction to the network device. Correspondingly, the terminal device receives the termination instruction from the network device.

In addition, for the scenario where one of the second neural network and the third neural network is a modulation network (the modulation network is used by the transmitter to process the data to be sent), and the other neural network is a detection network (the detection network is used by the receiver to process the received data), based on the information transmission method provided in the embodiment of the present application (the information transmission method as shown in Figure 8), the following configuration is used for simulation: the simulation channel is a Rayleigh flat fading channel, the modulation order is 4 bits/symbol (bits/symbol), the number of training samples is 100,000, and the performance indicator is the bit error rate (bit error rate, Ber). When the noise distribution is Gaussian (Gauss, GAU) noise, the simulation results of the information transmission method provided in the embodiment of the present application can be shown in Figure 15. When the noise distribution is a mixture of Gaussian (mixture of Gauss, MoG) noise, the simulation results of the information transmission method provided in the embodiment of the present application can be shown in Figure 16, wherein the mixed Gaussian noise contains two Gaussian components, the means of the two Gaussian components are -3 and 3, respectively, and the ratio of the two Gaussian components is 0.3:0.7.

As can be seen from Figures 15 and 16, compared with methods (1) to (4), the information transmission method provided by the embodiment of the present application has a lower bit error rate, thereby improving the signal decoding performance and the communication quality. Among them, method (1): traditional link (Trad_16QAM), the transmitting end uses 16 quadrature amplitude modulation (quadrature amplitude modulation, QAM) to modulate the signal to be transmitted, and the receiving end uses least squares (least square, LS) for channel estimation and linear minimum mean square error (linear minimum mean square error, LMMSE) for signal detection. Method (2): distribution matching intelligent link (AI_16QAM), the modulation network and the detection network are trained on a matching distribution data set. Method (3): distribution mismatched intelligent link (AI_XX_16QAM, when the noise distribution is mixed Gaussian noise, it is specifically AI_GAU_16QAM, when the noise distribution is mixed Gaussian noise, it is specifically AI_MoG_16QAM), the modulation network and the detection network are trained on a mismatched distribution data set. Method (4): Hybrid trained intelligent link (AImix_16QAM), which trains the modulation network and the detection network on a mixed dataset of two distributions.

In order to implement the functions of the method provided in the above embodiment of the present application, the first device or the second device may include a hardware structure and/or a software module, and implement the above functions in the form of a hardware structure, a software module, or a hardware structure plus a software module. Whether one of the above functions is executed in the form of a hardware structure, a software module, or a hardware structure plus a software module depends on the specific application and design constraints of the technical solution.

As shown in Figure 17, an embodiment of the present application provides a communication device 1700. The communication device 1700 can be a first device or a second device, or can also be a component of the first device (for example, an integrated circuit, a chip, etc.), or can also be a component of the second device (for example, an integrated circuit, a chip, etc.). The communication device 1700 can also be other communication units for implementing the method in the method embodiment of the present application. The communication device 1700 may include a processing unit 1701. Optionally, the communication device 1700 may also include a communication unit 1702, and the processing unit 1701 is used to control the communication unit 1702 to send and receive data/signaling, and the communication unit 1702 may also be referred to as a transceiver unit. Optionally, the communication unit 1702 may include a sending unit and a receiving unit, and the sending unit can be used to send data/signaling, and the receiving unit can be used to receive data/signaling. Optionally, the communication device 1700 may further include a storage unit 1703 , which may be used to store information and/or data and/or instructions, etc. The storage unit 1703 may interact with the processing unit 1701 , and may also interact with the communication unit 1702 .

In a possible design, for a case where the communication device 1700 is used to implement the function of the first device in the above method embodiment:

Processing unit 1701 is used to determine first adjustment information based on the first neural network, where the first adjustment information is used to indicate adjustment information of at least one neural network layer in the second neural network, and the second neural network is a neural network in the second device.

The communication unit 1702 is used to send first adjustment information to the second device, where the first adjustment information is used by the second device to adjust at least one neural network layer in the second neural network.

In an optional implementation, the first adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structure adjustment information of at least one neural network layer in the second neural network.

In an optional implementation, the first adjustment information is obtained by inputting information used to characterize the distribution of channel characteristics into the first neural network and outputting the information. Alternatively, the first adjustment information is obtained by inputting information used to characterize the distribution of channel characteristics and structural information of at least one neural network layer in the second neural network into the first neural network and outputting the information.

Optionally, the information used to characterize the channel characteristic distribution includes one or more of the following: channel state information, noise distribution information, signal-to-noise ratio, delay distribution information, power distribution information, Doppler spread information, and RSRP.

In an optional embodiment, the communication unit 1702 is further used to receive first indication information from the second device, where the first indication information is used to apply for adjusting the second neural network, or the first indication information is used to request activation of the first neural network.

In another optional implementation, the communication unit 1702 is further used to send a first indication message to the second device, where the first indication message is used to apply for adjusting the second neural network, or the first indication message is used to request activation of the first neural network.

In an optional implementation, the communication unit 1702 is further used to send second indication information to the second device, where the second indication information is used to indicate at least one neural network layer to be adjusted in the second neural network.

In another optional implementation, the communication unit 1702 is further used to receive second indication information from a second device, where the second indication information is used to indicate at least one neural network layer to be adjusted in the second neural network.

In an optional embodiment, processing unit 1701 is further used to determine second adjustment information based on the first neural network, where the second adjustment information is used to indicate adjustment information of at least one neural network layer in the third neural network; processing unit 1701 is further used to adjust at least one neural network layer in the third neural network based on the second adjustment information.

Optionally, the second adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structure adjustment information of at least one neural network layer in the third neural network.

In another possible design, for the case where the communication device 1700 is used to implement the function of the second device in the above method embodiment:

Communication unit 1702 is used to receive first adjustment information from the first device, where the first adjustment information is used to indicate adjustment information of at least one neural network layer in the second neural network, and the first adjustment information is determined by the first device based on the first neural network.

Processing unit 1701 is used to adjust at least one neural network layer in the second neural network based on the first adjustment information.

In an optional implementation, the first adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structure adjustment information of at least one neural network layer in the second neural network.

In an optional implementation, the first adjustment information is obtained by the first device inputting information used to characterize the distribution of channel characteristics into the first neural network. Alternatively, the first adjustment information is obtained by the first device inputting information used to characterize the distribution of channel characteristics and structural information of at least one neural network layer in the second neural network into the first neural network.

Optionally, the information used to characterize the channel characteristic distribution includes one or more of the following: channel state information, noise distribution information, signal-to-noise ratio, delay distribution information, power distribution information, Doppler spread information, and RSRP.

In an optional implementation, the communication unit 1702 is further used to send first indication information to the first device, where the first indication information is used to apply for adjusting the second neural network, or the first indication information is used to request activation of the first neural network.

In another optional embodiment, the communication unit 1702 is further used to receive first indication information from the first device, where the first indication information is used to apply for adjusting the second neural network, or the first indication information is used to request activation of the first neural network.

In an optional implementation, the communication unit 1702 is used to receive second indication information from the first device, where the second indication information is used to indicate at least one neural network layer to be adjusted in the second neural network.

In another optional implementation, the communication unit 1702 is further used to send second indication information to the first device, where the second indication information is used to indicate at least one neural network layer to be adjusted in the second neural network.

The application embodiment and the method embodiment shown above are based on the same concept, and the technical effects they bring are also the same. For the specific principles, please refer to the description of the embodiment shown above, which will not be repeated here.

The embodiment of the present application also provides a communication device 1800, as shown in Figure 18. The communication device 1800 can be a first device or a second device, or a chip, a chip system, or a processor that supports the first device to implement the above method, or a chip, a chip system, or a processor that supports the second device to implement the above method. The device can be used to implement the method described in the above method embodiment, and the details can be referred to the description in the above method embodiment.

The communication device 1800 may include one or more processors 1801. The processor 1801 may be used to implement part or all of the functions of the first device or the second device through a logic circuit or running a computer program. The processor 1801 may be a general-purpose processor or a dedicated processor, etc. For example, it may be a baseband processor, a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component or a central processing unit (CPU). The baseband processor may be used to process the communication protocol and the communication data, and the central processing unit may be used to control the communication device, execute the software program, and process the data of the software program, wherein the communication device is, for example, a base station, a baseband chip, a terminal, a terminal chip, a distributed unit (DU) or a distributed unit (CU), etc.

Optionally, the communication device 1800 may include one or more memories 1802, on which instructions 1804 may be stored, and the instructions may be executed on the processor 1801, so that the communication device 1800 performs the method described in the above method embodiment. Optionally, data may also be stored in the memory 1802. The processor 1801 and the memory 1802 may be provided separately or integrated together.

The memory 1802 may include, but is not limited to, non-volatile memories such as a hard disk drive (HDD) or a solid-state drive (SSD), random access memory (RAM), erasable programmable ROM (EPROM), ROM or portable read-only memory (compact disc read-only memory, CD-ROM), etc.

Optionally, the communication device 1800 may further include a transceiver 1805 and an antenna 1806. The transceiver 1805 may be referred to as a transceiver unit, a transceiver, or a transceiver circuit, etc., for implementing a transceiver function. The transceiver 1805 may include a receiver and a transmitter, the receiver may be referred to as a receiver or a receiving circuit, etc., for implementing a receiving function; the transmitter may be referred to as a transmitter or a transmitting circuit, etc., for implementing a transmitting function.

In a possible design, for a case where the communication device 1800 is used to implement the function of the first device in the above method embodiment:

Processor 1801 is used to determine first adjustment information based on a first neural network, where the first adjustment information is used to indicate adjustment information of at least one neural network layer in a second neural network, and the second neural network is a neural network in a second device.

Transceiver 1805 is used to send first adjustment information to the second device, where the first adjustment information is used by the second device to adjust at least one neural network layer in the second neural network.

In an optional implementation, the first adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structure adjustment information of at least one neural network layer in the second neural network.

In an optional implementation, the first adjustment information is obtained by inputting information used to characterize the distribution of channel characteristics into the first neural network and outputting the information. Alternatively, the first adjustment information is obtained by inputting information used to characterize the distribution of channel characteristics and structural information of at least one neural network layer in the second neural network into the first neural network and outputting the information.

Optionally, the information used to characterize the channel characteristic distribution includes one or more of the following: channel state information, noise distribution information, signal-to-noise ratio, delay distribution information, power distribution information, Doppler spread information, and RSRP.

In an optional embodiment, the transceiver 1805 is further used to receive first indication information from the second device, where the first indication information is used to apply for adjusting the second neural network, or the first indication information is used to request activation of the first neural network.

In another optional implementation, the transceiver 1805 is further used to send a first indication message to the second device, where the first indication message is used to apply for adjusting the second neural network, or the first indication message is used to request activation of the first neural network.

In an optional implementation, the transceiver 1805 is further used to send second indication information to the second device, where the second indication information is used to indicate at least one neural network layer to be adjusted in the second neural network.

In another optional embodiment, the transceiver 1805 is further used to receive second indication information from a second device, where the second indication information is used to indicate at least one neural network layer to be adjusted in the second neural network.

In an optional embodiment, processor 1801 is further used to determine second adjustment information based on the first neural network, where the second adjustment information is used to indicate adjustment information of at least one neural network layer in a third neural network; processor 1801 is further used to adjust at least one neural network layer in the third neural network based on the second adjustment information.

Optionally, the second adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structure adjustment information of at least one neural network layer in the third neural network.

In another possible design, for the case where the communication device 1800 is used to implement the function of the second device in the above method embodiment:

Transceiver 1805 is used to receive first adjustment information from the first device, where the first adjustment information is used to indicate adjustment information of at least one neural network layer in the second neural network, and the first adjustment information is determined by the first device based on the first neural network.

Processor 1801 is used to adjust at least one neural network layer in the second neural network based on the first adjustment information.

In an optional implementation, the first adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structure adjustment information of at least one neural network layer in the second neural network.

In an optional implementation, the first adjustment information is obtained by the first device inputting information used to characterize the distribution of channel characteristics into the first neural network. Alternatively, the first adjustment information is obtained by the first device inputting information used to characterize the distribution of channel characteristics and structural information of at least one neural network layer in the second neural network into the first neural network.

Optionally, the information used to characterize the channel characteristic distribution includes one or more of the following: channel state information, noise distribution information, signal-to-noise ratio, delay distribution information, power distribution information, Doppler spread information, and RSRP.

In an optional implementation, the transceiver 1805 is further used to send a first indication message to the first device, where the first indication message is used to apply for adjusting the second neural network, or the first indication message is used to request activation of the first neural network.

In another optional embodiment, the transceiver 1805 is further used to receive first indication information from the first device, where the first indication information is used to apply for adjusting the second neural network, or the first indication information is used to request activation of the first neural network.

In an optional embodiment, the transceiver 1805 is used to receive second indication information from the first device, and the second indication information is used to indicate at least one neural network layer to be adjusted in the second neural network.

In another optional implementation, the transceiver 1805 is further used to send second indication information to the first device, where the second indication information is used to indicate at least one neural network layer to be adjusted in the second neural network.

In another possible design, the processor 1801 may include a transceiver for implementing the receiving and sending functions. For example, the transceiver may be a transceiver circuit, or an interface, or an interface circuit. The transceiver circuit, interface, or interface circuit for implementing the receiving and sending functions may be separate or integrated. The above-mentioned transceiver circuit, interface, or interface circuit may be used for reading and writing code/data, or the above-mentioned transceiver circuit, interface, or interface circuit may be used for transmitting or delivering signals.

In another possible design, optionally, the processor 1801 may store an instruction 1803, and the instruction 1803 runs on the processor 1801, so that the communication device 1800 can execute the method described in the above method embodiment. The instruction 1803 may be solidified in the processor 1801, in which case the processor 1801 may be implemented by hardware.

In another possible design, the communication device 1800 may include a circuit that can implement the functions of sending or receiving or communicating in the aforementioned method embodiments. The processor and transceiver described in the embodiments of the present application can be implemented in an integrated circuit (IC), an analog IC, a radio frequency integrated circuit (RFIC), a mixed signal IC, an application specific integrated circuit (ASIC), a printed circuit board (PCB), an electronic device, etc. The processor and transceiver can also be manufactured using various IC process technologies, such as complementary metal oxide semiconductor (CMOS), N-type metal oxide semiconductor (nMetal-oxide-semiconductor, NMOS), P-type metal oxide semiconductor (positive channel metal oxide semiconductor, PMOS), bipolar junction transistor (bipolar junction transistor, BJT), bipolar CMOS (BiCMOS), silicon germanium (SiGe), gallium arsenide (GaAs), etc.

Those skilled in the art may also understand that the various illustrative logical blocks and steps listed in the embodiments of the present application may be implemented by electronic hardware, computer software, or a combination of the two. Whether such functions are implemented by hardware or software depends on the specific application and the design requirements of the entire system. Those skilled in the art may use various methods to implement the functions described for each specific application, but such implementation should not be understood as exceeding the scope of protection of the embodiments of the present application.

The embodiments of the present application and the above-mentioned method embodiments are based on the same concept, and the technical effects they bring are also the same. For the specific principles, please refer to the description in the above-mentioned method embodiments, which will not be repeated here.

The present application also provides a computer-readable storage medium for storing computer software instructions, which, when executed by a communication device, implement the functions of any of the above method embodiments.

The present application also provides a computer program product for storing computer software instructions, which, when executed by a communication device, implement the functions of any of the above method embodiments.

The present application also provides a computer program, which, when executed on a computer, implements the functions of any of the above method embodiments.

In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a magnetic tape), an optical medium (e.g., a high-density digital video disc (DVD)), or a semiconductor medium (e.g., an SSD), etc.

The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims (22)

An information transmission method, characterized in that the method comprises: Determining first adjustment information based on the first neural network, the first adjustment information being used to indicate adjustment information of at least one neural network layer in a second neural network, the second neural network being a neural network in a second device; The first adjustment information is sent to the second device, where the first adjustment information is used by the second device to adjust the at least one neural network layer in the second neural network. The method according to claim 1, characterized in that The first adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structure adjustment information of at least one neural network layer in the second neural network. The method according to claim 1 or 2, characterized in that The first adjustment information is obtained by inputting information used to characterize channel feature distribution into the first neural network and outputting the information; or, The first adjustment information is obtained by inputting information used to characterize channel feature distribution and structural information of at least one neural network layer in the second neural network into the output of the first neural network. The method according to claim 3, characterized in that The information used to characterize the channel characteristic distribution includes one or more of the following: channel state information, noise distribution information, signal-to-noise ratio, delay distribution information, power distribution information, Doppler spread information, and reference signal received power RSRP. The method according to any one of claims 1 to 4, characterized in that before determining the first adjustment information based on the first neural network, the method further comprises: Receive first indication information from the second device, where the first indication information is used to apply for adjusting the second neural network, or the first indication information is used to request activating the first neural network. The method according to any one of claims 1 to 5, characterized in that the method further comprises: Sending second indication information to the second device, where the second indication information is used to indicate the at least one neural network layer to be adjusted in the second neural network. The method according to any one of claims 1 to 4, characterized in that before determining the first adjustment information based on the first neural network, the method further comprises: Sending first indication information to the second device, where the first indication information is used to apply for adjusting the second neural network, or the first indication information is used to request activating the first neural network. The method according to any one of claims 1 to 4 and 7, characterized in that the method further comprises: Receive second indication information from the second device, where the second indication information is used to indicate the at least one neural network layer to be adjusted in the second neural network. The method according to any one of claims 1 to 8, characterized in that the method further comprises: Based on the first neural network, determining second adjustment information, wherein the second adjustment information is used to indicate adjustment information of at least one neural network layer in the third neural network; Based on the second adjustment information, at least one neural network layer in the third neural network is adjusted. The method according to claim 9, characterized in that The second adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structure adjustment information of at least one neural network layer in the third neural network. An information transmission method, characterized in that the method comprises: Receive first adjustment information from a first device, wherein the first adjustment information is used to indicate at least one neural network layer in a second neural network adjustment information, wherein the first adjustment information is determined by the first device based on the first neural network; Based on the first adjustment information, adjust the at least one neural network layer in the second neural network. The method according to claim 11, characterized in that The first adjustment information is specifically used to indicate one or more of the following: weight adjustment information, bias adjustment information, activation function, and structure adjustment information of at least one neural network layer in the second neural network. The method according to claim 11 or 12, characterized in that The first adjustment information is obtained by the first device inputting information used to characterize channel feature distribution into the first neural network and outputting it; or, The first adjustment information is obtained by the first device inputting information used to characterize channel feature distribution and structural information of at least one neural network layer in the second neural network into the first neural network and outputting it. The method according to claim 13, characterized in that The information used to characterize the channel characteristic distribution includes one or more of the following: channel state information, noise distribution information, signal-to-noise ratio, delay distribution information, power distribution information, Doppler spread information, and reference signal received power RSRP. The method according to any one of claims 11 to 14, characterized in that before receiving the first adjustment information from the first device, the method further comprises: Sending first indication information to the first device, where the first indication information is used to apply for adjusting the second neural network, or the first indication information is used to request activating the first neural network. The method according to any one of claims 11 to 15, characterized in that the method further comprises: Receive second indication information from the first device, where the second indication information is used to indicate the at least one neural network layer to be adjusted in the second neural network. The method according to any one of claims 11 to 14, characterized in that before receiving the first adjustment information from the first device, the method further comprises: Receive first indication information from the first device, where the first indication information is used to apply for adjusting the second neural network, or the first indication information is used to request activating the first neural network. The method according to any one of claims 11 to 14 and 17, characterized in that the method further comprises: Sending second indication information to the first device, where the second indication information is used to indicate the at least one neural network layer to be adjusted in the second neural network. A communication device, characterized in that the device includes a module or unit for implementing the method described in any one of claims 1 to 10, or the device includes a module or unit for implementing the method described in any one of claims 11 to 18. A communication device, comprising a processor; The processor is used to execute a computer program or instruction in a memory so that the communication device executes the method according to any one of claims 1 to 10, or so that the communication device executes the method according to any one of claims 11 to 18. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, and when the computer program is executed, it implements the method according to any one of claims 1 to 10, or implements the method according to any one of claims 11 to 18. A computer program product, characterized in that the computer program product comprises: computer program code, which, when the computer program code is run, implements the method according to any one of claims 1 to 10, or implements the method according to any one of claims 11 to 18.