一、 矩阵求导方法
- 对于映射 f : ℜ m ∗ n ↦ ℜ f:\Re^{m*n} \mapsto \Re f:ℜm∗n↦ℜ,即将 m ∗ n m * n m∗n的矩阵 A A A映射为实数 f ( A ) f(A) f(A),则函数 f f f关于 A A A的偏导为:
∇ A f ( A ) = [ ∂ f ∂ A 11 ⋯ ∂ f ∂ A 1 n ⋮ ⋱ ⋮ ∂ f ∂ A m 1 ⋯ ∂ f ∂ A m n ] \nabla_Af(A) = \left[ \begin{matrix} \frac{\partial f}{\partial A_{11}} & \cdots & \frac{\partial f}{\partial A_{1n} } \\ \vdots & \ddots & \vdots \\ \frac{\partial f}{\partial A_{m1} } & \cdots & \frac{\partial f}{\partial A_{mn} } \\ \end{matrix} \right] ∇Af(A)=⎣⎢⎡∂A11∂f⋮∂Am1∂f⋯⋱⋯∂A1n∂f⋮∂Amn∂f⎦⎥⎤ - t r tr tr trace operator
对于 n ∗ n n*n n∗n的方阵 A A A,则 T r ( A ) Tr(A) Tr(A)为A矩阵的对角线元素之和,即:
t r ( A ) = ∑ i = 1 n A i i tr(A) = \sum_{i=1}^{n} A_{ii} tr(A)=i=1∑nAii
实数的trace等于实数本身,即 t r ( a ) = a , a ∈ ℜ tr(a) = a,a \in \Re tr(a)=a,a∈ℜ
3.使用 t r tr tr对矩阵求偏导,公式如下: