本文为《Linear algebra and its applications》的读书笔记
目录
Eigenvectors and Eigenvalues
Eigenvectors and Eigenvalues
- Although a transformation x ↦ A x \boldsymbol x \mapsto A\boldsymbol x x↦Ax may move vectors in a variety of directions, it often happens that there are special vectors which are transformed by A A A into scalar multiples of themselves.
- Note that an eigenvector must be nonzero, by definition, but an eigenvalue may be zero.
- Let λ \lambda λ be an eigenvalue of an invertible matrix A A A. Then λ − 1 \lambda^{-1} λ−1 is an eigenvalue of A − 1 A^{-1} A−1.
Eigenspace
特征空间
-
λ
\lambda
λ is an eigenvalue of an
n
×
n
n \times n
n×n matrix
A
A
A if and only if the equation
( A − λ I ) x = 0 ( 3 ) (A-\lambda I)\boldsymbol x=\boldsymbol 0\ \ \ \ \ \ \ \ \ \ \ (3) (A−λI)x=0 (3)has a nontrivial solution. The set of all solutions of (3) is just the null space of the matrix A − λ I A-\lambda I A−λI. So this set is a subspace of R n \mathbb R^n Rn and is called the eigenspace (特征空间) of A A A corresponding to λ \lambda λ.- The eigenspace consists of the zero vector and all the eigenvectors corresponding to λ \lambda λ.
- Note that the linear combination of the eigenvectors corresponding to λ \lambda λ is still an eigenvector corresponding to λ \lambda λ.
EXAMPLE 3
Show that 7 is an eigenvalue of matrix A = [ 1 6 5 2 ] A=\begin{bmatrix}1&6\\5&2\end{bmatrix} A=[1562], and find the corresponding eigenvectors.
SOLUTION
- The scalar 7 is an eigenvalue of
A
A
A if and only if the equation
A x = 7 x A\boldsymbol x=7\boldsymbol x Ax=7xhas a nontrivial solution, which is equivalent to ( A − 7 I ) x = 0 (A-7I)\boldsymbol x=\boldsymbol 0 (A−7I)x=0 - To solve this homogeneous equation, form the matrix
The general solution has the form x 2 [ 1 1 ] x_2\begin{bmatrix} 1\\1\end{bmatrix} x2[11]. Each vector of this form with x 2 ≠ 0 x_2 \neq 0 x2=0 is an eigenvector corresponding to λ = 7 \lambda = 7 λ=7.
EXAMPLE 4
- Let
A
=
[
4
−
1
6
2
1
6
2
−
1
8
]
A=\begin{bmatrix}4&-1&6\\2&1&6\\2&-1&8\end{bmatrix}
A=⎣⎡422−11−1668⎦⎤. An eigenvalue of
A
A
A is 2. It can be shown that the basis for the corresponding eigenspace is
{
[
1
2
0
]
,
[
−
3
0
1
]
}
\{\begin{bmatrix}1\\2\\0\end{bmatrix},\begin{bmatrix}-3\\0\\1\end{bmatrix}\}
{⎣⎡120⎦⎤,⎣⎡−301⎦⎤}. So the eigenspace is a two-dimensional subspace of
R
3
\mathbb R^3
R3.
推论: A A A 与 A T A^T AT 的特征值相同
- Show that λ \lambda λ is an eigenvalue of A A A if and only if λ \lambda λ is an eigenvalue of A T A^T AT.
SOLUTION
- For any λ \lambda λ, ( A – λ I ) T = A T – ( λ I ) T = A T – λ I (A –\lambda I)^T = A^T – (\lambda I)^T = A^T – \lambda I (A–λI)T=AT–(λI)T=AT–λI. Since ( A – λ I ) T (A – \lambda I)^T (A–λI)T is invertible if and only if ( A – λ I ) (A – \lambda I) (A–λI) is invertible, we conclude that A T – λ I A^T – \lambda I AT–λI is not invertible if and only if A – λ I A – \lambda I A–λI is not invertible. That is, λ \lambda λ is an eigenvalue of A T A^T AT if and only if λ \lambda λ is an eigenvalue of A A A.
EXERCISES 30
Consider an n × n n \times n n×n matrix A A A with the property that the column sums all equal the same number s s s. Show that s s s is an eigenvalue of A A A.
SOLUTION
- [Hint: the row sums of
A
T
A^T
AT all equal the same number
s
s
s]
- 也可以证明 d e t ( A − s I ) = 0 det(A-sI)=0 det(A−sI)=0
Eigenvalues of a triangular matrix
- The following theorem describes one of the few special cases in which eigenvalues can be found precisely. Calculation of eigenvalues will also be discussed in Section 5.2.
PROOF
- For simplicity, consider the
3
×
3
3 \times 3
3×3 case. If
A
A
A is upper triangular, then
A
−
λ
I
A -\lambda I
A−λI has the form
The scalar λ \lambda λ is an eigenvalue of A A A if and only if the equation ( A − λ I ) x = 0 (A-\lambda I)\boldsymbol x=\boldsymbol 0 (A−λI)x=0 has a nontrivial solution, that is, if and only if the equation has a free variable. It is easy to see that ( A − λ I ) x = 0 (A-\lambda I)\boldsymbol x=\boldsymbol 0 (A−λI)x=0 has a free variable if and only if at least one of the entries on the diagonal of A − λ I A-\lambda I A−λI is zero. This happens if and only if λ \lambda λ equals one of the entries a 11 , a 22 , a 33 a_{11}, a_{22}, a_{33} a11,a22,a33 in A A A.
Eigenvalue of 0
- What does it mean for a matrix
A
A
A to have an eigenvalue of 0? This happens if and only if the equation
A x = 0 x A\boldsymbol x=0\boldsymbol x Ax=0xhas a nontrivial solution. But (4) is equivalent to A x = 0 A\boldsymbol x=\boldsymbol 0 Ax=0, which has a nontrivial solution if and only if A A A is not invertible. - Thus 0 is an eigenvalue of A A A if and only if A A A is not invertible. This fact will be added to the Invertible Matrix Theorem.
不同特征值对应的特征向量之间线性无关
PROOF
- Suppose
{
v
1
,
.
.
.
,
v
r
}
\{\boldsymbol v_1,..., \boldsymbol v_r\}
{v1,...,vr} is linearly dependent. Since
v
1
\boldsymbol v_1
v1 is nonzero, One of the vectors in the set is a linear combination of the preceding vectors. Let
p
p
p be the least index such that
v
p
+
1
\boldsymbol v_{p+1}
vp+1 is a linear combination of the preceding (linearly independent) vectors. Then there exist scalars
c
1
,
.
.
.
,
c
p
c_1,..., c_p
c1,...,cp such that
c 1 v 1 + . . . + c p v p = v p + 1 ( 5 ) c_1\boldsymbol v_1+...+ c_p\boldsymbol v_p=\boldsymbol v_{p+1}\ \ \ \ \ \ (5) c1v1+...+cpvp=vp+1 (5)Multiplying both sides of (5) by A A A, we obtain
c 1 A v 1 + . . . + c p A v p = A v p + 1 c 1 λ 1 v 1 + . . . + c p λ p v p = λ p + 1 v p + 1 ( 6 ) c_1A\boldsymbol v_1+...+ c_pA\boldsymbol v_p=A\boldsymbol v_{p+1}\\c_1\lambda_1\boldsymbol v_1+...+ c_p\lambda_p\boldsymbol v_p=\lambda_{p+1}\boldsymbol v_{p+1}\ \ \ \ \ \ (6) c1Av1+...+cpAvp=Avp+1c1λ1v1+...+cpλpvp=λp+1vp+1 (6)Multiplying both sides of (5) by λ p + 1 \lambda_{p+1} λp+1 and subtracting the result from (6), we have
c 1 ( λ 1 − λ p + 1 ) v 1 + . . . + c p ( λ p − λ p + 1 ) v p = 0 ( 7 ) c_1(\lambda_1-\lambda_{p+1})\boldsymbol v_1+...+ c_p(\lambda_p-\lambda_{p+1})\boldsymbol v_p=\boldsymbol 0\ \ \ \ \ \ (7) c1(λ1−λp+1)v1+...+cp(λp−λp+1)vp=0 (7) - Since { v 1 , . . . , v p } \{\boldsymbol v_1,..., \boldsymbol v_p\} {v1,...,vp} is linearly independent, the weights in (7) are all zero, which is impossible. Hence { v 1 , . . . , v r } \{\boldsymbol v_1,..., \boldsymbol v_r\} {v1,...,vr} cannot be linearly dependent and therefore must be linearly independent.
推论: n × n n \times n n×n 矩阵最多只有 n n n 个不同特征值
- Show that an n × n n \times n n×n matrix can have at most n n n distinct eigenvalues.
PROOF
- If an n × n n \times n n×n matrix has p p p distinct eigenvalues, then by Theorem 2 there would be a linearly independent set of p p p eigenvectors (one for each eigenvalue). Since these vectors belong to an n n n-dimensional vector space, p p p cannot exceed n n n.
Eigenvectors and Difference Equations
- This section concludes by showing how to construct solutions of the first-order difference equation:
x k + 1 = A x k ( k = 0 , 1 , 2 , . . . ) ( 8 ) \boldsymbol x_{k+1}=A\boldsymbol x_k\ \ \ \ \ \ (k=0,1,2,...)\ \ \ \ \ \ \ \ (8) xk+1=Axk (k=0,1,2,...) (8) - If A A A is an n × n n \times n n×n matrix, then (8) is a recursive description of a sequence { x k } \{\boldsymbol x_k\} {xk} in R n \mathbb R^n Rn. A solution of (8) is an explicit description of { x k } \{\boldsymbol x_k\} {xk} whose formula for each x k \boldsymbol x_k xk does not depend directly on A A A or on the preceding terms in the sequence other than the initial term x 0 \boldsymbol x_0 x0.
- The simplest way to build a solution of (8) is to take an eigenvector
x
0
\boldsymbol x_0
x0 and its corresponding eigenvalue
λ
\lambda
λ and let
x k = λ k x 0 ( k = 1 , 2 , . . . ) ( 9 ) \boldsymbol x_{k}=\lambda^k\boldsymbol x_0\ \ \ \ \ \ (k=1,2,...)\ \ \ \ \ \ \ \ (9) xk=λkx0 (k=1,2,...) (9)This sequence is a solution because
A x k = A ( λ k x 0 ) = λ k ( A x 0 ) = λ k ( λ x 0 ) = λ k + 1 x 0 = x k + 1 A\boldsymbol x_{k}=A(\lambda^k\boldsymbol x_{0})=\lambda^k(A\boldsymbol x_{0})=\lambda^k(\lambda \boldsymbol x_{0})=\lambda^{k+1}\boldsymbol x_{0}=\boldsymbol x_{k+1} Axk=A(λkx0)=λk(Ax0)=λk(λx0)=λk+1x0=xk+1Linear combinations of solutions in the form of equation (9) are solutions, too!