Repository navigation
Expand file tree
/
Copy pathalge.tex
More file actions
675 lines (560 loc) · 15 KB
/
Copy pathalge.tex
File metadata and controls
675 lines (560 loc) · 15 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
10BC8
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
\documentclass{article}
\usepackage{amsmath}
\author{a1trl9}
\title{Algebra Review}
\date{}
\setlength\parindent{0pt}
\counterwithin*{equation}{section}
\newcommand{\V}[1]{\boldsymbol{#1}}
\newcommand{\TM}{\bigtriangledown}
\begin{document}
\maketitle
\section{System of Linear Equations}
\textbf{Proposition 1:} for a homogeneous system of linear equations on
\(K\), if the number of equations \(m\) is fewer than the number of
variables \(n\), then it must have a non-trival solution.
\vspace{2mm}
\textbf{Mathematical Induction}
When \(m=1, n>m\), if one coefficient \(a_{1k}=0\), then one
solution is \(x_k=1\) and
all other variables are equal to \(0\). If no
coefficient is equal to \(0\). Then one solution is
\(x_k = -\frac{a_{1k+1}}{a_{k}},
x_{k+1}=1\), all other variables are equal to \(0\).
\vspace{1mm}
Considering when \(m > 1\) while for \(m-1\) the proposition holds.
If all coefficients for \(x_1\) are equal to zero, then one solution is
\(x_1=1, x_2=0, \cdots, x_n=0\).
If not, then change the order of equations so that \(\alpha_{11} \neq 0\).
Now repeat the step \(eq(k) - \frac{a_{11}}{a_{k1}}eq(1)\) where \(eq(t)\)
presents the equation \(t\) and \(k=2,3,\cdots,n\). Finally, the system
of equations are transformed to:
\begin{equation}
\left\{
\begin{split}
&a_{11}x_1+a_{12}x_2+\cdots+a_{1n}x_n=0 \\
&a'_{22}x_2 + \cdots + a'_{2n}x_n=0 \\
&\vdots\\
&a'_{m2}x_2 + \cdots + a'_{mn}x_n=0 \\
\end{split}
\right.
\end{equation}
Now the second to last equations construct a new homogeneous system
of equations with \(m-1\) equations and \(n-1\) variables:
\begin{equation}
\left\{
\begin{split}
&a'_{22}x_2 + \cdots + a'_{2n}x_n=0 \\
&a'_{32}x_2 + \cdots + a'_{3n}x_n=0 \\
&\vdots\\
&a'_{m2}x_2 + \cdots + a'_{mn}x_n=0 \\
\end{split}
\right.
\end{equation}
Since \(n>m\), \(n-1>m-1\). And the proposition hold for \(m-1\). There
exists one non-trival solution \(x_2=k_2, \cdots, x_n=k_n\).
Back to (1). Since \(a_{11} \neq 0\), with
\(x_2=k_2, \cdots, x_n=k_n\), there must be one solution \(x_1=k_1\)
letting \(a_{11}k_1+a_{12}k_2+\cdots+a_{1n}k_n=0\).
\section{Vector space \& Matrix \& System of Linear Equations}
\subsection{Vector space}
\textbf{Proposition 2.1.4:} Given two sets of vectors:
\begin{equation}
\begin{split}
\alpha_1, \alpha_2, \cdots, \alpha_r, \quad\quad (\mathbf{I}) \\
\beta_1, \beta, \cdots, \beta_s, \quad\quad (\mathbf{II})
\end{split}
\end{equation}
If each vector in \((\mathbf{I})\) can be defined as a linear combination
of \((\mathbf{II})\), and \(r>s\). Then \((\mathbf{I})\) are linearly dependent.
\vspace{2mm}
\textbf{Proof}
As each vector in \((\mathbf{I})\) can be defined as a linear combination
of \((\mathbf{II})\):
\begin{equation}
\begin{split}
&\alpha_1 = x_{11}\beta_1 + x_{12}\beta_2 + \cdots + x_{1s}\beta_s \\
&\alpha_2 = x_{21}\beta_1 + x_{22}\beta_2 + \cdots + x_{2s}\beta_s \\
&\vdots\\
&\alpha_r = x_{r1}\beta_1 + x_{r2}\beta_2 + \cdots + x_{rs}\beta_s \\
\end{split}
\end{equation}
Then given \(k_1, k_2, \cdots, k_s\):
\begin{equation}
\sum_{i=1}^{r}k_i\alpha_i=
\sum_{j=1}^{s}x_{j1}k_j\beta_1
+\sum_{j=1}^{s}x_{j2}k_j\beta_2
+\cdots
+\sum_{j=1}^{s}x_{jr}k_j\beta_r
\end{equation}
In which, since \(r>s\), for:
\begin{equation}
\left
\{
\begin{split}
&x_{11}k_1 + x_{21}k_2 + \cdots + x_{s1}k_s = 0 \\
&x_{12}k_1 + x_{22}k_2 + \cdots + x_{s2}k_s = 0 \\
&\vdots\\
&x_{1r}k_1 + x_{2r}k_2 + \cdots + x_{sr}k_s = 0 \\
\end{split}
\right.
\end{equation}
According to proposition 1 in section 1, there
must be a non-trival solution. That is, there exists \(k_1, k_2, \cdots, k_s\),
in which at least one \(k_n \neq 0\), for \(i = 1, 2, \cdots, r\), \(\sum_{j=1}^s
x_{ji}k_j=0\), in which case \(\sum_{i=1}^rk_i\alpha_i=0\) as well.
. So, \(\alpha_1, \alpha_2,
\cdots, \alpha_r\) are linearly dependent.
\subsection{Matrix operations}
\subsubsection{Operations and rank}
\textbf{Proposition 2.2.4:}: \(r(AB)\leq min(r(A), r(B))\).
\vspace{2mm}
\textit{Lemma 1:}
For two matrics:
\begin{equation}
\begin{split}
&A=\begin{bmatrix}
\alpha_1, \alpha_2, \cdots, \alpha_r \\
\end{bmatrix}\\
&B=\begin{bmatrix}
\beta_1, \beta_2, \cdots, \beta_s
\end{bmatrix}
\end{split}
\end{equation}
if each column vector of \(A\) could be defined as a linear combination
of column vectors in \(B\), then \(r(A)\leq r(B)\).
\vspace{2mm}
\textbf{Proof}
Since each column vector of \(A\) could be defined as a linear combination
of column vectors in \(B\), the maximally linearly independent set
of \(A\): \(\alpha_{i_1}, \alpha_{i_2}, \cdots, \alpha_{i_k}\)
could be defined as a linear combination of
the maximally linearly independent set of \(B\):
\(\beta_{j_1}, \beta_{j_2}, \cdots, \beta_{j_m}\) as well.
Since \(\alpha_{i_1}, \alpha_{i_2}, \cdots, \alpha_{i_k}\) must be
linearly independent, according to proposition 2.1.4,
\(i_k \leq j_m \), that is \(r(A)\leq r(B)\).
\vspace{2mm}
\textbf{Proof for proposition 2.2.4}
Let \(C=AB=\begin{bmatrix}c_1,c_2,\cdots,c_n\end{bmatrix}\). Then
for \(c_i, i = 1, 2, \cdots, n\):
\begin{equation}
\begin{split}
c_i &= \begin{bmatrix}
a_{11}b_{i1}+a_{12}b_{i2}+\cdots+a_{1m}b_{im}\\
a_{21}b_{i1}+a_{22}b_{i2}+\cdots+a_{2m}b_{im}\\
a_{k1}b_{i1}+a_{k2}b_{i2}+\cdots+a_{km}b_{im}\\
\end{bmatrix}\\
&=\alpha_1b_{i1}+\alpha_2b_{i2}+\cdots+\alpha_mb_{im}
\end{split}
\end{equation}
That is, each column vector of \(C\) could be presented as
a combination of column vectors of \(A\). So according
to the lemma, \(r(C)\leq r(A)\). Similarly, \(r(C)\leq r(B)\)
(using transpose). Therefore, \(r(AB)\leq min(r(A), r(B))\).
\subsection{Square matrix}
\textbf{Proposition 2.3.3:} for a \(n\) square matrix \(A\) in the number
field \(K\), the sufficient and necessary condition for \(A\) is
invertable is \(A\) has full rank.
\vspace{2mm}
\textbf{Proof}
\vspace{2mm}
\textit{Sufficiency}
When \(A\) has full rank, there exist elementary matrics
\(P_i, Q_j, i = 1, 2, \cdots, n, j=1,2,\cdots,m\), such that:
\begin{equation}
P_1P_2\cdots P_nA = E
\end{equation}
And:
\begin{equation}
AQ_1Q_2\cdots Q_m = E
\end{equation}
Let \(P=P_1P_2\cdots P_n\), and \(Q=Q_1Q_2\cdots Q_m\).
Then:
\begin{equation}
P=PE=PAQ=(PA)Q=EQ=Q
\end{equation}
So \(AP=PA=E\), A is invertable.
\vspace{2mm}
\textit{Necessity}
Since \(A\) is invertable, there exists \(P\), \(AP=E\).
According to proposition 2.2.4, \(r(E)=n\leq min(r(A), r(P))\leq r(A)\).
Therefore \(r(A)=n\) and has full rank.
\vspace{2mm}
\textbf{Proposition 2.3.5}: For a \(n \times n\) square matrix \(\mathbf{X}\),
\(X"X\) is a positive semi-definite matrix.
\vspace{2mm}
\textbf{Proof}
If we present \(X=[X_1, X_2, \cdots, X_n]'\) and
\(X'=[X_1, X_2, \cdots, X_n]\)
It can be shown \(\mathbf{A}=\mathbf{X}'\mathbf{X}\) is symmetric as
\begin{equation}
a_{ij}=X_i \cdot X_j =a_{ji}
\end{equation}
Where \(a_{kl}\) presents the element located in the \(k\)th row and \(l\)th column
of \(A\).
Then for any non-zero vector \(\V{v}\):
\begin{equation}
\V{v}'A\V{v}=\V{v}'\mathbf{X}'\mathbf{X}\V{v}=(\mathbf{X}\V{v})'\mathbf{X}\V{v}
=||\mathbf{X}\V{v}||^2\geq 0
\end{equation}
\section{Determinant}
\subsection{Definition}
\subsubsection{Linearity and Antisymmetry}
Considering the set consisting of all
\(n \times n\) matrix \(M_n(K)\) in the number field \(K\) and
a function \(f\) in the domain,
the linearity by row embodies as:
For matrix:
\begin{equation*}
A=
\begin{bmatrix}
\alpha_1 \\
\alpha_2 \\
\vdots \\
\alpha_n
\end{bmatrix}
\end{equation*}
If for \(i \in [1, n]\):
\begin{equation*}
\alpha_i=k\alpha+l\beta
\end{equation*}
Then:
\begin{equation}
f(A)=
\begin{bmatrix}
\alpha_1 \\
\vdots \\
\alpha_i \\
\vdots \\
\alpha_n
\end{bmatrix} =
kf(
\begin{bmatrix}
\alpha_1 \\
\vdots \\
\alpha \\
\vdots \\
\alpha_n
\end{bmatrix}) + lf(
\begin{bmatrix}
\alpha_1 \\
\vdots \\
\beta \\
\vdots \\
\alpha_n
\end{bmatrix}
)
\end{equation}
Similarly, the linearity by column embodies as:
\begin{equation}
\begin{split}
\begin{bmatrix}
\alpha_1 &
\cdots &
\alpha_i &
\cdots &
\alpha_n
\end{bmatrix} & =
kf(
\begin{bmatrix}
\alpha_1 &
\cdots &
\alpha &
\cdots &
\alpha_n
\end{bmatrix}) \\
& + lf(
\begin{bmatrix}
\alpha_1 &
\cdots &
\beta &
\cdots &
\alpha_n
\end
10BC8
{bmatrix}
)
\end{split}
\end{equation}
Moreover, the function \(f\) holding the characteristic above is
said antisymmetric if
\(f(A)=0\) when any two rows/columns of \(A\) are identical.
\vspace{2mm}
\textbf{Proposition 3.1.1:}
if function \(f\) holds all characteristics above (i.e. linear
by row/column, antisymmetric), then:
\begin{itemize}
\item Swap two rows/coloums of matrix \(A\) to matrix \(B\),
\(f(A)=-f(B)\).
\item Add \(\lambda\alpha_j\) to \(\alpha_i\) to get matrix \(B\),
\(f(A)=f(B)\).
\end{itemize}
\vspace{2mm}
\textbf{Proof:}
Assuming \(f\) holds linearity by row, then:
\begin{equation}
\begin{split}
f(
\begin{bmatrix}
\alpha_1 \\
\vdots \\
\alpha_i \\
\vdots \\
\alpha_j \\
\vdots \\
\alpha_n
\end{bmatrix}) +
f(\begin{bmatrix}
\alpha_1 \\
\vdots \\
\alpha_j \\
\vdots \\
\alpha_i \\
\vdots \\
\alpha_n
\end{bmatrix})
=
f(\begin{bmatrix}
\alpha_1 \\
\vdots \\
\alpha_i + \alpha_j \\
\vdots \\
\alpha_j + \alpha_i \\
\vdots \\
\alpha_n
\end{bmatrix})=0
\end{split}
\end{equation}
Therefore:
\begin{equation}
f(
\begin{bmatrix}
\alpha_1 \\
\vdots \\
\alpha_i \\
\vdots \\
\alpha_j \\
\vdots \\
\alpha_n
\end{bmatrix}) =
-f(\begin{bmatrix}
\alpha_1 \\
\vdots \\
\alpha_j \\
\vdots \\
\alpha_i \\
\vdots \\
\alpha_n
\end{bmatrix})
\end{equation}
Besides:
\begin{equation}
f(
\begin{bmatrix}
\alpha_1 \\
\vdots \\
\alpha_i + \lambda\alpha_j \\
\vdots \\
\alpha_j \\
\vdots \\
\alpha_n
\end{bmatrix}) =
f(\begin{bmatrix}
\alpha_1 \\
\vdots \\
\alpha_i \\
\vdots \\
\alpha_j \\
\vdots \\
\alpha_n
\end{bmatrix}) +
\lambda f(\begin{bmatrix}
\alpha_1 \\
\vdots \\
\alpha_j \\
\vdots \\
\alpha_j \\
\vdots \\
\alpha_n
\end{bmatrix}) =
f(\begin{bmatrix}
\alpha_1 \\
\vdots \\
\alpha_i \\
\vdots \\
\alpha_j \\
\vdots \\
\alpha_n
\end{bmatrix}) + 0 =
f(\begin{bmatrix}
\alpha_1 \\
\vdots \\
\alpha_i \\
\vdots \\
\alpha_j \\
\vdots \\
\alpha_n
\end{bmatrix})
\end{equation}
For \(f\) holding the linearity by column, the proof is similar.
\vspace{2mm}
\textbf{Inference 1}
Together with the rule indicated in (1), namely:
\begin{equation}
f(\begin{bmatrix}
\alpha_1 \\
\vdots \\
\lambda\alpha_i \\
\vdots \\
\alpha_n
\end{bmatrix})=
\lambda f(\begin{bmatrix}
\alpha_1 \\
\vdots \\
\alpha_i \\
\vdots \\
\alpha_n
\end{bmatrix})
\end{equation}
And:
\begin{equation}
f(\begin{bmatrix}
\alpha_1 &
\cdots &
\lambda\alpha_i &
\cdots &
\alpha_n
\end{bmatrix})=
\lambda f(\begin{bmatrix}
\alpha_1 &
\cdots &
\alpha_i &
\cdots &
\alpha_n
\end{bmatrix})
\end{equation}
These three rules correspond to three types of elementary transformations
by row/column. Therefore, it is obvious that for two functions \(f,g\) holding
antisymmetry and linearity, if \(f(A)=g(A)\) and \(B=AP_1P_2\cdots P_n\)
where \(P_1, P_2, \cdots, P_n\) are all elementary matrics, then
\(f(B)=g(B)\).
\vspace{2mm}
\textbf{Inference 2}
The necessary and sufficient condition for a function \(f\) holding
linearity by row/column to be antisymmetric is for each matrix
\(A\) that is rank deficient, \(f(A)=0\).
\vspace{2mm}
\textbf{Proof:}
\vspace{1mm}
\textit{Sufficient:}
Since for each matrix \(A\) that is rank deficient, \(f(A)=0\),
and obviously if \(A\) has two identical rows/columns, \(A\) is
rank deficient. So \(f(A)=0\) when \(A\) has two identical rows/columns.
\vspace{2mm}
\textit{Necessary:}
Assuming \(f\) holds linearity by row. Then for \(\lambda \neq 0\):
\begin{equation}
\begin{bmatrix}
\alpha_1 \\
\vdots \\
\mathbf{0} \\
\vdots \\
\alpha_n
\end{bmatrix} =
\begin{bmatrix}
\alpha_1 \\
\vdots \\
\lambda \mathbf{0} \\
\vdots \\
\alpha_n
\end{bmatrix} =
\lambda\begin{bmatrix}
\alpha_1 \\
\vdots \\
\mathbf{0} \\
\vdots \\
\alpha_n
\end{bmatrix}
\end{equation}
Therefore:
\begin{equation}
\begin{bmatrix}
\alpha_1 \\
\vdots \\
\mathbf{0} \\
\vdots \\
\alpha_n
\end{bmatrix} = 0
\end{equation}
Besides, if \(A\) is rank deficient, with \(\alpha_i = k_1\alpha_1 + \cdots +
k_{i-1}\alpha_{i-1} + k_{i+1}\alpha_{i+1} + \cdots + k_n\alpha_n\),
repeat the step that \(\alpha'_i=\alpha_i - k_j \alpha_{j}\) for
\(k = 1, 2, \cdots, i-1, i+1, \cdots, n\). Finally, \(\alpha'_i=\mathbf{0}\).
Considering \(f\) holds the linearity:
\begin{equation}
f(A)=f(
\begin{bmatrix}
\alpha_1 \\
\vdots \\
\alpha_i \\
\vdots \\
\alpha_n
\end{bmatrix})=
f(\begin{bmatrix}
\alpha_1 \\
\vdots \\
\alpha'_i \\
\vdots \\
\alpha_n
\end{bmatrix})=
f(\begin{bmatrix}
\alpha_1 \\
\vdots \\
\mathbf{0} \\
\vdots \\
\alpha_n
\end{bmatrix})=0
\end{equation}
For \(f\) holding the linearity by column, the proof is similar.
\subsubsection{Definition}
There are three characteristics that a determinant function holds:
\begin{itemize}
\item If matrix \(A\) is rank deficient, \(f(A)=0\).
\item For identity matrix \(E\), \(f(E)=1\).
\item Linearity by column.
\end{itemize}
\vspace{2mm}
\textbf{Proposition 3.1.2:} The determinant function is unique.
\vspace{2mm}
\textbf{Proof:}
Assuming there are two determinant functions: \(f, g\).
For matrix \(A\), When \(r(A)<n\), according to the three characteristics,
\(f(A)=g(A)=0\).
When \(r(A)=n\), \(A\) can be presented as \(EP_1P_2\cdots P_m\) where
\(E\) is an identity matrix and \(P_1, P_2, \cdots, P_m\) are elementary
matrics. As \(f(E)=g(E)\), together with \textbf{Inference 1},
\(f(A)=g(A)\).
So \(f, g\) are identical.
\vspace{2mm}
\textbf{Proposition 3.1.3:}
For a determinant function \(f\), \(f(A)=f(A')\) where \(A'\) is
the transpose of matrix \(A\).
\vspace{2mm}
\textbf{Proof:}
When \(r(A)<n\), then \(f(A)=f(A')=0\).
When \(r(A)=n\), \(A\) can be presented as \(EP_1P_2\cdots P_m\) where
\(E\) is an identity matrix and \(P_1, P_2, \cdots, P_m\) are elementary
matrics. \(f(A)=(-1)^r\lambda_1\lambda_2\cdots \lambda_s\) where \(r,s\)
are numbers of the type I \& type II elementary matrics (i.e. \(P_i=P_n(j,k)\) or
\(P_i=P_n(\lambda\cdot k)\)) respectively.
Similarly, \(A'=EP'_mP'_{m-1}\cdots P'_1\).
\(f(A')=(-1)^{r'}\lambda_1\lambda_2\cdots \lambda_{s'}\).
When \(P_i\) is the
first or second type of elementary, matrics , \(P_i=P'_i\). So \(P_1P_2\cdots P_m\)
and \(P'_mP'_{m-1}\cdots P'_1\) contain the same number of type I and type
II elementary matrics.
When \(P_i\) is the
third type (i.e. \(P_i=P_n(\lambda\cdot j, k)\)), \(P'_i\) is also the
third type, therefore \(f(XP_i)=f(X)\) and \(f(XP'_i)=f(X)\) (\(P_i,
P'_i\) will not change the output of \(f\)).
So \(f(A)=f(EP_1P_2\cdots P_m)=f(EP'_mP'_{m-1}\cdots P'_1)=f(A')\).
\end{document}