本文为《Linear algebra and its applications》的读书笔记
目录
The inverse of a matrix
- Recall that the multiplicative inverse (乘法逆) of a number such as 5 is 1/5 or
5
−
1
5^{-1}
5−1. This inverse satisfies the equations
5 − 1 ⋅ 5 = 1 , 5 ⋅ 5 − 1 = 1 5^{-1}\cdot5=1,\ \ \ \ \ \ 5\cdot5^{-1}=1 5−1⋅5=1, 5⋅5−1=1 - The matrix generalization requires b o t h both both equations and avoids the slanted-line (斜线) notation (for division) because matrix multiplication is not commutative (可交换的).
- Furthermore, a full generalization is possible only if the matrices involved are square. (Check Example 1~3)
EXAMPLE 1
Suppose C A = I n CA = I_n CA=In (the n × n n \times n n×n identity matrix). Show that A A A cannot have more columns than rows.
SOLUTION
- If x \boldsymbol x x satisfies A x = 0 A\boldsymbol x = \boldsymbol 0 Ax=0, then C A x = C 0 = 0 CA\boldsymbol x = C\boldsymbol 0 = \boldsymbol 0 CAx=C0=0 and so I n x = 0 I_n\boldsymbol x = \boldsymbol 0 Inx=0 and x = 0 \boldsymbol x = \boldsymbol 0 x=0. This implies that A x = 0 A\boldsymbol x = \boldsymbol 0 Ax=0 only has the trivial solution and the echelon form of A A A must have no free variables. Thus A A A cannot have more columns than rows.
EXAMPLE 2
Suppose A D = I m AD = I_m AD=Im. Show that A A A cannot have more rows than columns.
SOLUTION
- For any b \boldsymbol b b in R m \mathbb R^m Rm, the equation A x = b Ax = \boldsymbol b Ax=b has a solution. ( A D b = b AD\boldsymbol b = \boldsymbol b ADb=b) Thus A A A cannot have more rows than columns.
- We can also apply the transpose operation to obtain D T A T = I m D^TA^T=I_m DTAT=Im. Then the result from example 1 can lead us to the conclusion.
EXAMPLE 3
Suppose A A A is an m × n m \times n m×n matrix and there exist n × m n \times m n×m matrices C C C and D D D such that C A = I n CA = I_n CA=In and A D = I m AD = I_m AD=Im: Prove that m = n m = n m=n and C = D C = D C=D.
SOLUTION
- By EXAMPLE 1, the equation C A = I n CA = I_n CA=In implies that m ≥ n m\geq n m≥n. By EXAMPLE 2, the equation A D = I m AD = I_m AD=Im implies that m ≤ n m\leq n m≤n. Thus m = n m=n m=n.
- To prove the second statement, C = C I n = C ( A D ) = ( C A ) D = I n D = D C = CI_n = C(AD) = (CA)D = I_nD = D C=CIn=C(AD)=(CA)D=InD=D
代数结构中,若运算满足结合律且左右逆元均存在,则左右逆元相同 ( C = D C=D C=D)。同时,逆元唯一,即 A A A 的逆唯一
- An
n
×
n
n \times n
n×n matrix
A
A
A is said to be invertible (可逆的) if there is an
n
×
n
n \times n
n×n matrix
C
C
C such that
C A = I n , A C = I n CA=I_n,\ \ \ \ AC=I_n CA=In, AC=InIn this case, C C C is an inverse (逆) of A A A. C C C is uniquely determined by A A A. This unique inverse is denoted by A − 1 A^{-1} A−1, so that
A − 1 A = I n , A A − 1 = I n A^{-1}A=I_n,\ \ \ \ AA^{-1}=I_n A−1A=In, AA−1=In - A matrix that is not invertible is sometimes called a singular matrix (奇异矩阵), and an invertible matrix is called a nonsingular matrix (非奇异矩阵).
已知 A A A 为 n × n n\times n n×n 矩阵,且 C A = I CA=I CA=I,则 A x = 0 A\boldsymbol x =\boldsymbol 0 Ax=0 只有平凡解, A A A 有 n n n 个主元列, A ∼ I A\sim I A∼I,因此 A A A 可逆, C = A − 1 C=A^{-1} C=A−1
同理已知 A A A 为 n × n n\times n n×n 矩阵,且 A D = I AD=I AD=I,则 A A A 可逆, D = A − 1 D=A^{-1} D=A−1
- Here is a simple formula for the inverse of a 2 × 2 2 \times 2 2×2 matrix, along with a test to tell if the inverse exists.
- The quantity
a
d
−
b
c
ad - bc
ad−bc is called the determinant (行列式) of
A
A
A, and we write
d e t A = a d − b c det\ A = ad - bc det A=ad−bc - Theorem 4 says that a 2 × 2 2 \times 2 2×2 matrix A A A is invertible if and only if d e t A ≠ 0 det\ A \neq 0 det A=0.
- Numerical Notes:
- The formula in Theorem 5 is seldom used to solve an equation A x = b A\boldsymbol x = \boldsymbol b Ax=b numerically because row reduction of [ A b ] [\ \ A\ \ \ \boldsymbol b\ \ ] [ A b ] is nearly always faster. (Row reduction is usually more accurate, too, when computations involve rounding off numbers.)
- One possible exception is the 2 × 2 2 \times 2 2×2 case. In this case, mental computations to solve A x = b A\boldsymbol x = \boldsymbol b Ax=b are sometimes easier using the formula for A − 1 A^{-1} A−1, as in the next example.
EXAMPLE 4
Use the inverse of the matrix to solve the system
SOLUTION
- The following generalization of Theorem 6(b) is needed later.
Elementary Matrices
初等矩阵
- There is an important connection between invertible matrices and row operations that leads to a method for computing inverses. As we shall see, an invertible matrix A A A is row equivalent to an identity matrix, and we can find A − 1 A^{-1} A−1 by watching the row reduction of A A A to I I I.
- An elementary matrix is one that is obtained by performing a single elementary row operation on an identity matrix.
- The next example illustrates the three kinds of elementary matrices.
EXAMPLE 5
Let
Compute
E
1
A
E_1A
E1A,
E
2
A
E_2A
E2A, and
E
3
A
E_3A
E3A, and describe how these products can be obtained by elementary row operations on
A
A
A.
SOLUTION
- Addition of
−
4
-4
−4 times row 1 of
A
A
A to row 3 produces
E
1
A
E_1A
E1A. (This is a row replacement operation.) An interchange of rows 1 and 2 of
A
A
A produces
E
2
A
E_2A
E2A, and multiplication of row 3 of
A
A
A by 5 produces
E
3
A
E_3A
E3A.
- Left-multiplication by E 1 E_1 E1 in Example 5 has the same effect on any 3 × n 3 \times n 3×n matrix. It adds − 4 -4 −4 times row 1 to row 3. In particular, since E 1 ⋅ I = E 1 E_1 \cdot I = E_1 E1⋅I=E1, we see that E 1 E_1 E1 itself is produced by this same row operation on the identity.
- Example 5 illustrates the following general fact about elementary matrices.
PROOF (下面的证明以
3
×
3
3\times3
3×3 矩阵为例)
∵
A
=
I
⋅
A
∴
r
o
w
i
(
A
)
=
r
o
w
i
(
I
)
⋅
A
\begin{aligned} \because A&=I\cdot A \\\therefore row_i(A)&=row_i(I)\cdot A \end{aligned}
∵A∴rowi(A)=I⋅A=rowi(I)⋅Aa. Show that if rows 1 and 2 of
A
A
A are interchanged, then the result may be written as
E
A
EA
EA, where
E
E
E is an elementary matrix formed by interchanging rows 1 and 2 of
I
I
I.
b. Show that if row 3 of
A
A
A is multiplied by 5, then the result may be written as
E
A
EA
EA, where
E
E
E is formed by multiplying row 3 of
I
I
I by 5.
- Since row operations are reversible, elementary matrices are invertible. Hence there is an elementary matrix F F F such that F E = I FE = I FE=I. Since E E E and F F F correspond to reverse operations, E F = I EF = I EF=I, too.
EXAMPLE 6
Find the inverse of
E
1
=
[
1
0
0
0
1
0
−
4
0
1
]
E_1=\begin{bmatrix}1&0&0\\0&1&0\\-4&0&1\end{bmatrix}
E1=⎣⎡10−4010001⎦⎤.
SOLUTION
E
1
−
1
=
[
1
0
0
0
1
0
4
0
1
]
E_1^{-1}=\begin{bmatrix}1&0&0\\0&1&0\\4&0&1\end{bmatrix}
E1−1=⎣⎡104010001⎦⎤
- The following theorem provides the best way to “visualize” an invertible matrix, and the theorem leads immediately to a method for finding the inverse of a matrix.
PROOF
- Suppose that A A A is invertible. Then, since the equation A x = b A\boldsymbol x =\boldsymbol b Ax=b has a solution for each b \boldsymbol b b, A A A has a pivot position in every row. Because A A A is square, the n n n pivot positions must be on the diagonal, which implies that the reduced echelon form of A A A is I n I_n In. That is, A ∼ I n A \sim I_n A∼In.
- Now suppose, conversely, that
A
∼
I
n
A \sim I_n
A∼In. Then, since each step of the row reduction of
A
A
A corresponds to left-multiplication by an elementary matrix, there exist elementary matrices
E
1
,
.
.
.
,
E
p
E_1,...,E_p
E1,...,Ep such that
A ∼ E 1 A ∼ E 2 ( E 1 A ) ∼ . . . ∼ E p ( E p − 1 . . . E 1 A ) = I n A \sim E_1A \sim E_2(E_1A)\sim ...\sim E_p(E_{p-1}...E_1A) = I_n A∼E1A∼E2(E1A)∼...∼Ep(Ep−1...E1A)=InThat is,
E p . . . E 1 A = I n ( 1 ) E_p...E_1A = I_n\ \ \ \ \ \ \ \ (1) Ep...E1A=In (1) Since the product E p . . . E 1 E_p...E_1 Ep...E1 of invertible matrices is invertible, ( 1 ) (1) (1) leads to
( E p . . . E 1 ) − 1 ( E p . . . E 1 ) A = ( E p . . . E 1 ) − 1 I n A = ( E p . . . E 1 ) − 1 \begin{aligned}(E_p...E_1)^{-1}(E_p...E_1)A &= (E_p...E_1)^{-1}I_n\\A&=(E_p...E_1)^{-1}\end{aligned} (Ep...E1)−1(Ep...E1)AA=(Ep...E1)−1In=(Ep...E1)−1Thus A A A is invertible, as it is the inverse of an invertible matrix (Theorem 6). Also,
A − 1 = [ ( E p . . . E 1 ) − 1 ] − 1 = E p . . . E 1 A^{-1}=[(E_p...E_1)^{-1}]^{-1}= E_p...E_1 A−1=[(Ep...E1)−1]−1=Ep...E1Then A − 1 = E p . . . E 1 ⋅ I n A^{-1}= E_p...E_1\cdot I_n A−1=Ep...E1⋅In, which says that A − 1 A^{-1} A−1 results from applying E 1 , . . . , E p E_1,...,E_p E1,...,Ep successively to I n I_n In. This is the same sequence in (1) that reduced A A A to I n I_n In.
An Algorithm for Finding A – 1 A^{–1} A–1
EXAMPLE 7
Find the inverse of the matrix
A
=
[
0
1
2
1
0
3
4
−
3
8
]
A=\begin{bmatrix} 0&1&2\\1&0&3\\4&-3&8\end{bmatrix}
A=⎣⎡01410−3238⎦⎤, if it exists.
SOLUTION
举一反三:
- 如果想计算
A
−
1
B
A^{-1}B
A−1B 的话,也可以由行化简求得:若
[
A
B
]
∼
.
.
.
∼
[
I
X
]
[A\ \ \ B]\sim ...\sim[I\ \ \ X]
[A B]∼...∼[I X],则
X
=
A
−
1
B
X=A^{-1}B
X=A−1B
- 如果 A A A 是大于 2 × 2 2\times2 2×2 的矩阵,则用行化简的方法求要比直接求快得多
- We can even accelerate this process by using the LU factorization of A A A, which is adapted by MATLAB.
Another View of Matrix Inversion
- Denote the columns of
I
n
I_n
In by
e
1
,
.
.
.
,
e
n
\boldsymbol e_1,...,\boldsymbol e_n
e1,...,en. Then row reduction of
[
A
I
]
[\ \ A \ \ \ I\ \ ]
[ A I ] to
[
I
A
−
1
]
[\ \ I \ \ \ A^{-1}\ \ ]
[ I A−1 ] can be viewed as the simultaneous solution of the
n
n
n systems
A x 1 = e 1 , A x 2 = e 2 , . . . , A x n = e n ( 2 ) A\boldsymbol x_1=\boldsymbol e_1,A\boldsymbol x_2=\boldsymbol e_2,...,A\boldsymbol x_n=\boldsymbol e_n\ \ \ \ \ (2) Ax1=e1,Ax2=e2,...,Axn=en (2)where the “augmented columns” of these systems have all been placed next to A A A to form [ A e 1 e 2 . . . e n ] = [ A I ] [\ \ A\ \ \boldsymbol e_1\ \ \boldsymbol e_2\ \ ... \ \ \boldsymbol e_n\ \ ] =[\ \ A\ \ I\ \ ] [ A e1 e2 ... en ]=[ A I ]. The equation A A − 1 = I AA^{-1} = I AA−1=I and the definition of matrix multiplication show that the columns of A − 1 A^{-1} A−1 are precisely the solutions of the systems in (2). - This observation is useful because some applied problems may require finding only one or two columns of A − 1 A^{-1} A−1. In this case, only the corresponding systems in (2) need be solved.
EXAMPLE 8
Let A = [ 1 0 . . . . . . . . . 0 1 1 0 . . . . . . 0 1 1 1 0 . . . 0 . . . . . . . . . . . . . . . . . . 1 1 1 1 1 0 1 1 1 1 1 1 ] A=\begin{bmatrix} 1&0&...&...&...&0\\ 1&1&0&...&...&0\\1&1&1&0&...&0\\ ...&...&...&...&...&... \\ 1&1&1&1&1&0\\ 1&1&1&1&1&1 \end{bmatrix} A=⎣⎢⎢⎢⎢⎢⎢⎡111...11011...11...01...11......0...11............11000...01⎦⎥⎥⎥⎥⎥⎥⎤. A A A is an n × n n\times n n×n matrix. Let B B B be its inverse. Guess the form of B B B, and then prove that A B = I AB = I AB=I and B A = I BA = I BA=I.
SOLUTION
- Let a j \boldsymbol a_j aj, b j \boldsymbol b_j bj, and e j \boldsymbol e_j ej denote the j j jth columns of A A A, B B B, and I I I, respectively. Note that for j = 1 , … , n – 1 j = 1, …, n – 1 j=1,…,n–1, a j – a j + 1 = e j \boldsymbol a_j – \boldsymbol a_{j+1} = \boldsymbol e_j aj–aj+1=ej.
- Suppose B = A − 1 B=A^{-1} B=A−1, then A b j = e j = a j − a j + 1 = A e j − A e j + 1 = A ( e j − e j + 1 ) A\boldsymbol b_j=\boldsymbol e_j=\boldsymbol a_j-\boldsymbol a_{j+1}=A\boldsymbol e_j-A\boldsymbol e_{j+1}=A(\boldsymbol e_j-\boldsymbol e_{j+1}) Abj=ej=aj−aj+1=Aej−Aej+1=A(ej−ej+1), A b n = e n = a n = A e n A\boldsymbol b_n=\boldsymbol e_n=\boldsymbol a_n=A\boldsymbol e_n Abn=en=an=Aen. Thus b j = e j – e j + 1 \boldsymbol b_j = \boldsymbol e_j – \boldsymbol e_{j+1} bj=ej–ej+1 and b n = e n \boldsymbol b_n = \boldsymbol e_n bn=en.
也可以从初等行变换的角度来考虑