常用矩阵求导
-
∂ β x ∂ x = β \frac{\partial \beta \mathbf{x}}{\partial \mathbf{x}}=\beta ∂x∂βx=β
-
∂ x T a ∂ x = ∂ a T x ∂ x = a \frac{\partial \mathbf{x}^T\mathbf{a}}{\partial \mathbf{x}}=\frac{\partial \mathbf{a}^T\mathbf{x}}{\partial \mathbf{x}}=\mathbf{a} ∂x∂xTa=∂x∂aTx=a
证明:
-
∂ x T x ∂ x = 2 x \frac{\partial \mathbf{x}^T\mathbf{x}}{\partial \mathbf{x}}=2\mathbf{x} ∂x∂xTx=2x
-
∂ x T A x ∂ x = ( A + A T ) x \frac{\partial \mathbf{x}^TA\mathbf{x}}{\partial \mathbf{x}}=(A+A^T)\mathbf{x} ∂x∂xTAx=(A+AT)x
-
t r ( a ) = a tr(a)=a tr(a)=a
-
t r ( A B ) = t r ( B A ) tr(AB)=tr(BA) tr(AB)=tr(BA)
-
t r ( A B C ) = t r ( C A B ) = t r ( B C A ) tr(ABC)=tr(CAB)=tr(BCA) tr(ABC)=tr(CAB)=tr(BCA)
-
t r ( A ) = t r ( A T ) tr(A)=tr(A^T) tr(A)=tr(AT)
-
∂ t r ( A B ) ∂ A = B T \frac{\partial tr(AB)}{\partial A}=B^T ∂A∂tr(AB)=BT
-
∂ t r ( A B A T C ) ∂ A = C A B + C T A B T \frac{\partial tr(ABA^TC)}{\partial A}=CAB+C^TAB^T ∂A∂tr(ABATC)=CAB+CTABT
逆矩阵求导
两种方法:
-
直接法
d X − 1 d x = lim Δ X → 0 ( X + Δ X ) − 1 − X − 1 Δ x = lim Δ X → 0 ( X + Δ X ) − 1 X X − 1 − ( X + Δ X ) − 1 ( X + Δ X ) X − 1 Δ x = lim Δ X → 0 ( X + Δ X ) − 1 X − ( X + Δ X ) Δ x X − 1 = lim Δ X → 0 ( X + Δ X ) − 1 − Δ X Δ x X − 1 = − ( X + Δ X ) − 1 lim Δ X → 0 Δ X Δ x X − 1 = − X − 1 lim Δ X → 0 Δ X Δ x X − 1 = − X − 1 d X d x X − 1 \begin{align}\frac{d X^{-1}}{dx}&=\lim_{\Delta X\rightarrow 0}\frac{(X+\Delta X)^{-1}-X^{-1}}{\Delta x}\\ &=\lim_{\Delta X\rightarrow 0}\frac{(X+\Delta X)^{-1}XX^{-1}-(X+\Delta X)^{-1}(X+\Delta X)X^{-1}}{\Delta x}\\ &=\lim_{\Delta X\rightarrow 0}(X+\Delta X)^{-1}\frac{X-(X+\Delta X)}{\Delta x}X^{-1}\\ &=\lim_{\Delta X\rightarrow 0}(X+\Delta X)^{-1}\frac{-\Delta X}{\Delta x}X^{-1}\\ &=-(X+\Delta X)^{-1}\lim_{\Delta X\rightarrow 0}\frac{\Delta X}{\Delta x}X^{-1}\\ &=-X^{-1}\lim_{\Delta X\rightarrow 0}\frac{\Delta X}{\Delta x}X^{-1}\\ &=-X^{-1}\frac{d X}{d x}X^{-1}\\ \end{align} dxdX−1=ΔX→0limΔx(X+ΔX)−1−X−1=ΔX→0limΔx(X+ΔX)−1XX−1−(X+ΔX)−1(X+ΔX)X−1=ΔX→0lim(X+ΔX)−1ΔxX−(X+ΔX)X−1=ΔX→0lim(X+ΔX)−1Δx−ΔXX−1=−(X+ΔX)−1ΔX→0limΔxΔXX−1=−X−1ΔX→0limΔxΔXX−1=−X−1dxdXX−1 -
利用 0 = ( X X − 1 ) ′ 0=(XX^{-1})^{'} 0=(XX−1)′
0 = ( X X − 1 ) ′ = X ′ X − 1 + X ( X − 1 ) ′ \begin{align}0&=(XX^{-1})^{'}=X^{'}X^{-1}+X(X^{-1})^{'} \end{align} 0=(XX−1)′=X′X−1+X(X−1)′
Then,
⇒ ( X − 1 ) ′ = − X − 1 X ′ X − 1 \Rightarrow (X^{-1})^{'}=-X^{-1}X^{'}X^{-1} ⇒(X−1)′=−X−1X′X−1