矩阵论复习笔记:矩阵直积及其应用

1.矩阵直积及其应用类题目

1.1 常用矩阵直积知识回顾

直积常用性质:

  • 1. A ⨂ B ≠ B ⨂ A A\bigotimes B \neq B\bigotimes A AB=BA

  • 2. ( A 1 + A 2 ) ⨂ B = A 1 ⨂ B + A 2 ⨂ B (A_1+A_2)\bigotimes B=A_1\bigotimes B+A_2\bigotimes B (A1+A2)B=A1B+A2B

  • 3. ( A ⨂ B ) ⨂ C = A ⨂ ( B ⨂ C ) (A\bigotimes B)\bigotimes C=A\bigotimes (B\bigotimes C) (AB)C=A(BC)

  • 4.若 A 1 , A 2 A_1,A_2 A1,A2可以做乘法运算, B 1 , B 2 B_1,B_2 B1,B2可以做乘法运算:
    ( A 1 ⨂ A 2 ) ( B 1 ⨂ B 2 ) = ( A 1 A 2 ) ⨂ ( B 1 B 2 ) (A_1\bigotimes A_2)(B_1\bigotimes B_2)=(A_1A_2)\bigotimes(B_1B_2) (A1A2)(B1B2)=(A1A2)(B1B2)

  • 5.若 A , B A,B A,B可以求逆:
    ( A ⨂ B ) − 1 = A − 1 ⨂ B − 1 (A \bigotimes B)^{-1}=A^{-1}\bigotimes B^{-1} (AB)1=A1B1
    若不能求逆运算则:
    ( A ⨂ B ) + = A + ⨂ B + (A\bigotimes B)^{+}=A^{+}\bigotimes B^{+} (AB)+=A+B+

  • 6. ( A ⨂ B ) H = A H ⨂ B H (A\bigotimes B)^H=A^H\bigotimes B^H (AB)H=AHBH

  • 7. d e t ( A ⨂ B ) = ( d e t A ) n ( d e t B ) m ( A ∈ C m × m , B ∈ C n × n ) det(A\bigotimes B)=(detA)^n(detB)^m(A\in C^{m\times m},B\in C^{n\times n}) det(AB)=(detA)n(detB)m(ACm×m,BCn×n)

  • 8. t r ( A ⨂ B ) = ( t r A ) ⨂ ( t r B ) tr(A\bigotimes B)=(trA)\bigotimes (trB) tr(AB)=(trA)(trB)

  • 9. r a n k ( A ⨂ B ) = r a n k A ⨂ r a n k B rank(A\bigotimes B)=rankA\bigotimes rankB rank(AB)=rankArankB

  • 10 e I ⨂ A = I ⨂ e A , e A ⨂ I = A ⨂ I e^{I \bigotimes A} = I\bigotimes e^A,e^{A\bigotimes I} = A\bigotimes I eIA=IeA,eAI=AI

  • 11. e ( A ⨂ I n + I m ⨂ B ) = e A ⨂ e B e^{(A\bigotimes I_n+I_m \bigotimes B)}=e^A\bigotimes e^B e(AIn+ImB)=eAeB

  • 12.设 f ( A , B ) = ∑ i = 1 l 1 ∑ j = 1 l 2 c i j A i ⨂ B j f(A,B)=\sum_{i=1}^{l_1}\sum_{j=1}^{l_2}c_{ij}A^{i}\bigotimes B^{j} f(A,B)=i=1l1j=1l2cijAiBj对应的二元多项式为; f ( λ , μ ) = ∑ i = 1 l 1 ∑ j = 1 l 2 c i j ( λ ) i ( μ ) j f(\lambda,\mu)=\sum_{i=1}^{l_1}\sum_{j=1}^{l_2}c_{ij}(\lambda)^i(\mu)^j f(λ,μ)=i=1l1j=1l2cij(λ)i(μ)j

    A A A的特征值为 λ 1 , λ 2 , . . . λ m \lambda_1,\lambda_2,...\lambda_m λ1,λ2,...λm, B B B的特征值为 μ 1 , μ 2 , . . . μ n \mu_1,\mu_2,...\mu_n μ1,μ2,...μn。得到的 f ( A , B ) f(A,B) f(A,B)的所有特征值为:
    f ( λ s , μ r ) = ∑ i = 1 l 1 ∑ j = 1 l 2 c i j λ s i μ r j s = 1 , 2.. m , r = 1 , 2.. n f(\lambda_s,\mu_r)=\sum_{i=1}^{l_1}\sum_{j=1}^{l_2}c_{ij}\lambda_s^i\mu_r^j\\ s=1,2..m,r = 1,2..n f(λs,μr)=i=1l1j=1l2cijλsiμrjs=1,2..m,r=1,2..n

1.2 线性矩阵方程可解性

A ∈ C m × p , X ∈ C p × q , B ∈ C q × n A\in C^{m\times p},X\in C^{p\times q},B\in C^{q\times n} ACm×p,XCp×q,BCq×n, v e c ˉ ( X ) \bar{vec}(X) vecˉ(X) X X X的行拉直向量,则有:
v e c ˉ ( A X B ) = ( A ⨂ B T ) v e c ˉ ( X ) \bar{vec}(AXB)=(A\bigotimes B^T)\bar{vec}(X) vecˉ(AXB)=(ABT)vecˉ(X)
A m × m A_{m\times m} Am×m的特征值为 λ 1 , λ 2 , . . . λ m \lambda_1,\lambda_2,...\lambda_m λ1,λ2,...λm B n × n B_{n\times n} Bn×n的特征值为 μ 1 , μ 2 , . . . μ n \mu_1,\mu_2,...\mu_n μ1,μ2,...μn,方程 ∑ k = 0 l A k X B k = F \sum_{k=0}^lA^kXB^k=F k=0lAkXBk=F有唯一解的充要条件是 ∀ λ i , μ j \forall \lambda_i,\mu_j λi,μj,有 ∑ k = 0 l ( λ i μ j ) k ≠ 0 \sum_{k=0}^l(\lambda_i\mu_j)^{k}\neq 0 k=0l(λiμj)k=0。方程 ∑ k = 0 l A k X B k = O \sum_{k=0}^lA^kXB^k=O k=0lAkXBk=O非零解的充要条件是 ∃ λ i , μ j \exists \lambda_i,\mu_j λi,μj,有 ∑ k = 0 l ( λ i μ j ) k = 0 \sum_{k=0}^l(\lambda_i\mu_j)^{k}= 0 k=0l(λiμj)k=0。同样的,对于以下的方程:
d Y ( t ) d t = A Y ( t ) + Y ( t ) B , Y ( t ) ∣ t = 0 = F \frac{dY(t)}{dt}=AY(t)+Y(t)B,Y(t)|_{t=0}=F dtdY(t)=AY(t)+Y(t)B,Y(t)t=0=F
A , B A,B A,B的特征值之和不等于零,
如果积分 X ∫ 0 ∞ e A t F e B t d t X \int_0^{\infty}e^{At}Fe^{Bt}dt X0eAtFeBtdt存在,则方程有唯一的解:

X = − ∫ 0 ∞ e A t F e B t d t X=-\int_0^{\infty}e^{At}Fe^{Bt}dt X=0eAtFeBtdt

若特征值实部 R e ( λ i ) < 0 , R e ( μ j ) < 0 Re(\lambda_i)<0,Re(\mu_j)<0 Re(λi)<0,Re(μj)<0,则 X X X有唯一的解 X = − ∫ 0 ∞ e A t F e B t d t X=-\int_0^{\infty}e^{At}Fe^{Bt}dt X=0eAtFeBtdt

换句话说求解得到该方程的唯一解为:
X = e A t X 0 e B t X=e^{At}X_0e^{Bt} X=eAtX0eBt

2.例题

n n n阶方阵 A A A的特征值为: λ 1 , λ 2 , . . . λ n \lambda_1,\lambda_2,...\lambda_n λ1,λ2,...λn,则矩阵方程 A X A 2 + X A − X = 0 AXA^2+XA-X=0 AXA2+XAX=0只有唯一零解的充分必要条件()。

解:对该矩阵方程进行行拉直向量变换:
v e c ˉ ( A X A 2 + X A − X ) = 0 \bar{vec}(AXA^2+XA-X)=0 vecˉ(AXA2+XAX)=0
​ 得到:
( A ⨂ ( A 2 ) T + I ⨂ A T − I ⨂ I ) v e c ˉ ( X ) = 0 (A\bigotimes (A^2)^T+I\bigotimes A^T -I\bigotimes I)\bar{vec}(X)=0 (A(A2)T+IATII)vecˉ(X)=0

f ( A , A T ) = ( A ⨂ ( A 2 ) T + I ⨂ A T − I ⨂ I ) f(A,A^T)=(A\bigotimes (A^2)^T+I\bigotimes A^T -I\bigotimes I) f(A,AT)=(A(A2)T+IATII)的二元多项式为:
f ( x , y ) = x y 2 + y − 1 f(x,y)=xy^2+y-1 f(x,y)=xy2+y1
若要使得上述线性方程组有唯一零解,即说明其特征值必须全部非零。即 ∀ μ i , μ j , f ( μ i , μ j ) ≠ 0 \forall \mu_i,\mu_j,f(\mu_i,\mu_j)\neq 0 μi,μj,f(μi,μj)=0

即:
μ i μ j 2 + μ j − 1 ≠ 0 ( i , j = 1 , 2.. n ) \mu_i\mu_j^2+\mu_j-1\neq 0(i,j=1,2..n) μiμj2+μj1=0(i,j=1,2..n)

你可能感兴趣的:(闲散杂记,矩阵,线性代数,机器学习)