向量化算子 V e c \mathbf{Vec} Vec
设 A = [ a i j ] m × n \mathbf{A=[a_{ij}]_{m\times n}} A=[aij]m×n,则有 V e c ( A ) = ( a 11 a 21 ⋯ a m 1 ; a 12 a 22 ⋯ a m 2 ; ⋯ ; a 1 n a 2 n ⋯ a m n ) ⊤ \mathbf{Vec(A)=(a_{11}a_{21}\cdots a_{m1};a_{12}a_{22}\cdots a_{m2};\cdots;a_{1n}a_{2n}\cdots a_{mn})}^{\top} Vec(A)=(a11a21⋯am1;a12a22⋯am2;⋯;a1na2n⋯amn)⊤
向量化算子性质
- 线性性质: V e c \mathbf{Vec} Vec是线性算子 V e c ( k 1 A + k 2 B ) = k 1 V e c ( A ) + k 2 V e c B \mathbf{Vec(k_1A+k_2B)=k_1Vec(A)+k_2Vec{B}} Vec(k1A+k2B)=k1Vec(A)+k2VecB
- 矩阵乘法: V e c ( A B C ) = ( C ⊤ ⊗ A ) V e c ( B ) \mathbf{Vec(ABC)=(C^{\top}\otimes A)Vec(B)} Vec(ABC)=(C⊤⊗A)Vec(B), V e c ( A X ) = ( I ⊗ A ) V e c ( X ) \mathbf{Vec(AX)=(I \otimes A)Vec(X)} Vec(AX)=(I⊗A)Vec(X), V e c ( X C ) = ( C ⊤ ⊗ I ) V e c ( X ) \mathbf{Vec(XC)=(C^{\top}\otimes I)Vec(X)} Vec(XC)=(C⊤⊗I)Vec(X)
- 矩阵转置: V e c ( A ⊤ ) = K m n V e c ( A ) \mathbf{Vec}(\mathbf{A}^{\top})=\mathbf{K}_{mn}\mathbf{Vec(A)} Vec(A⊤)=KmnVec(A)其中, K m n \mathbf{K}_{mn} Kmn是 m n × m n mn\times mn mn×mn的交换矩阵
- 逐元素乘法: V e c ( A ⊙ X ) = d i a g ( V e c ( A ) ) V e c ( X ) \mathbf{Vec(A\odot X)}=diag(\mathbf{Vec(A)})\mathbf{Vec(X)} Vec(A⊙X)=diag(Vec(A))Vec(X)其中 d i a g ( V e c ( A ) ) diag(\mathbf{Vec(A)}) diag(Vec(A))是 m n × m n mn \times mn mn×mn的对角矩阵。