矩阵求导之二：定义篇（上）

有了上一篇文章的直观感受，本篇文章将针对同类型的函数、变元给出严谨的导数定义。因为向量和标量都可以表示成简单的矩阵形式，所以这里我们使用矩阵来泛化的表示所有含义。规定：粗体小写字母（如$\mathbf{x}$）表示向量，粗体大写字母（如$\mathbf{X}$）表示矩阵。

1 标量对向量求导

1.1 定义

设$f\left(\mathbf{x}\right)=f\left(x_1,x_2,\cdots ,x_n\right)$，$\mathbf{x}={\left[x_1,x_2,\cdots ,x_n\right]}^T$，则：
在分子布局下：
$$\frac{\partial f}{\partial {\mathbf{x}}^T}=\left[\frac{\partial f}{\partial x_1},\frac{\partial f}{\partial x_2},\cdots ,\frac{\partial f}{\partial x_n}\right]$$
在分母布局下：
$$\frac{\partial f}{\partial \mathbf{x}}={\left[\frac{\partial f}{\partial x_1},\frac{\partial f}{\partial x_2},\cdots ,\frac{\partial f}{\partial x_n}\right]}^T$$

1.2 运算法则

1.2.1 线性法则

对于$f\left(\mathbf{x}\right)$，$g\left(\mathbf{x}\right)$，则：
$$\frac{\partial \left[c_{1}f\left(\mathbf{x}\right)\pm c_{2}g\left(\mathbf{x}\right)\right]}{\partial \mathbf{x}}=c_1\frac{\partial f\left(\mathbf{x}\right)}{\partial \mathbf{x}}\pm c_2\frac{\partial g\left(\mathbf{x}\right)}{\partial \mathbf{x}}$$
【证明】
$$\begin{array}{rl}\frac{\partial \left[c_{1}f\left(\mathbf{x}\right)\pm c_{2}g\left(\mathbf{x}\right)\right]}{\partial \mathbf{x}}& =\left[\begin{array}{c}\frac{\partial \left(c_{1}f+c_{2}g\right)}{\partial x_1}\\ \\ \frac{\partial \left(c_{1}f+c_{2}g\right)}{\partial x_2}\\ \\ ⋮\\ \\ \frac{\partial \left(c_{1}f+c_{2}g\right)}{\partial x_n}\end{array}\right]=\left[\begin{array}{c}{c}_1\frac{\partial f}{\partial x_1}+c_2\frac{\partial g}{\partial x_1}\\ \\ c_1\frac{\partial f}{\partial x_2}+c_2\frac{\partial g}{\partial x_2}\\ \\ ⋮\\ \\ c_1\frac{\partial f}{\partial x_n}+c_2\frac{\partial g}{\partial x_n}\end{array}\right]\\ & \\ & =c_1\left[\begin{array}{c}\frac{\partial f}{\partial x_1}\\ \\ \frac{\partial f}{\partial x_2}\\ \\ ⋮\\ \\ \frac{\partial f}{\partial x_n}\end{array}\right]+c_2\left[\begin{array}{c}\frac{\partial f}{\partial x_1}\\ \\ \frac{\partial f}{\partial x_2}\\ \\ ⋮\\ \\ \frac{\partial f}{\partial x_n}\end{array}\right]\\ & \\ & =c_1\frac{\partial f\left(\mathbf{x}\right)}{\partial \mathbf{x}}\pm c_2\frac{\partial g\left(\mathbf{x}\right)}{\partial \mathbf{x}}\end{array}$$

1.2.2 乘积公式

$$\frac{\partial \left[f\left(\mathbf{x}\right)g\left(\mathbf{x}\right)\right]}{\partial \mathbf{x}}=\frac{\partial f\left(\mathbf{x}\right)}{\partial \mathbf{x}}g\left(\mathbf{x}\right)\pm f\left(\mathbf{x}\right)\frac{\partial g\left(\mathbf{x}\right)}{\partial \mathbf{x}}$$
【证明】
$$\begin{array}{rl}\frac{\partial \left[f\left(\mathbf{x}\right)g\left(\mathbf{x}\right)\right]}{\partial \mathbf{x}}& =\left[\begin{array}{c}\frac{\partial \left(fg\right)}{\partial x_1}\\ \\ \frac{\partial \left(fg\right)}{\partial x_2}\\ \\ ⋮\\ \\ \frac{\partial \left(fg\right)}{\partial x_n}\end{array}\right]=\left[\begin{array}{c}\frac{\partial f}{\partial x_1}\cdot g+f\cdot \frac{\partial g}{\partial x_1}\\ \\ \frac{\partial f}{\partial x_2}\cdot g+f\cdot \frac{\partial g}{\partial x_2}\\ \\ ⋮\\ \\ \frac{\partial f}{\partial x_n}\cdot g+f\cdot \frac{\partial g}{\partial x_n}\end{array}\right]=\left[\begin{array}{c}\frac{\partial f}{\partial x_1}\\ \\ \frac{\partial f}{\partial x_2}\\ \\ ⋮\\ \\ \frac{\partial f}{\partial x_n}\end{array}\right]g+f\left[\begin{array}{c}\frac{\partial g}{\partial x_1}\\ \\ \frac{\partial g}{\partial x_2}\\ \\ ⋮\\ \\ \frac{\partial g}{\partial x_n}\end{array}\right]\\ & \\ & =\frac{\partial f\left(\mathbf{x}\right)}{\partial \mathbf{x}}g\left(\mathbf{x}\right)\pm f\left(\mathbf{x}\right)\frac{\partial g\left(\mathbf{x}\right)}{\partial \mathbf{x}}\end{array}$$

1.2.3 商公式

$$\frac{\partial \left[\frac{f\left(\mathbf{x}\right)}{g\left(\mathbf{x}\right)}\right]}{\partial \mathbf{x}}=\frac{1}{g^2\left(\mathbf{x}\right)}\left[\frac{\partial f\left(\mathbf{x}\right)}{\partial \mathbf{x}}g\left(\mathbf{x}\right)-f\left(\mathbf{x}\right)\frac{\partial g\left(\mathbf{x}\right)}{\partial \mathbf{x}}\right]$$

【证明】
$$\begin{array}{rl}\frac{\partial \left[\frac{f\left(\mathbf{x}\right)}{g\left(\mathbf{x}\right)}\right]}{\partial \mathbf{x}}& =\left[\begin{array}{c}\frac{\partial \left(f/g\right)}{\partial x_1}\\ \frac{\partial \left(f/g\right)}{\partial x_2}\\ ⋮\\ \frac{\partial \left(f/g\right)}{\partial x_n}\end{array}\right]=\left[\begin{array}{c}\frac{1}{g^2}\left(\frac{\partial f}{\partial x_1}g-f\frac{\partial g}{\partial x_1}\right)\\ \\ \frac{1}{g^2}\left(\frac{\partial f}{\partial x_2}g-f\frac{\partial g}{\partial x_2}\right)\\ \\ ⋮\\ \\ \frac{1}{g^2}\left(\frac{\partial f}{\partial x_n}g-f\frac{\partial g}{\partial x_n}\right)\end{array}\right]\\ & \\ & =\frac{1}{g^2}\left(\left[\begin{array}{c}\frac{\partial f}{\partial x_1}\\ \\ \frac{\partial f}{\partial x_2}\\ \\ ⋮\\ \\ \frac{\partial f}{\partial x_n}\end{array}\right]g-f\left[\begin{array}{c}\frac{\partial g}{\partial x_1}\\ \\ \frac{\partial g}{\partial x_2}\\ \\ ⋮\\ \\ \frac{\partial g}{\partial x_n}\end{array}\right]\right)\\ & \\ & =\frac{1}{g^2\left(\mathbf{x}\right)}\left[\frac{\partial f\left(\mathbf{x}\right)}{\partial \mathbf{x}}g\left(\mathbf{x}\right)-f\left(\mathbf{x}\right)\frac{\partial g\left(\mathbf{x}\right)}{\partial \mathbf{x}}\right]\end{array}$$
其中
$$g\left(\mathbf{x}\right)\ne 0$$

1.3 示例

【例1.1】设$\mathbf{x}={\left[x_1,x_2,\cdots ,x_m\right]}^T$,求函数$$f\left(\mathbf{x}\right)={\mathbf{x}}^T\mathbf{x}={x_1}^2+{x_2}^2+\cdots +{x_n}^2$$
对$\mathbf{x}$的导数。
【解】
$$\frac{\partial f}{\partial \mathbf{x}}=\left[\begin{array}{c}\frac{\partial f}{\partial x_1}\\ \\ \frac{\partial f}{\partial x_2}\\ \\ ⋮\\ \\ \frac{\partial f}{\partial x_n}\end{array}\right]=\left[\begin{array}{c}2x_1\\ 2x_2\\ ⋮\\ 2x_n\end{array}\right]=2\mathbf{x}$$
即：
$$\frac{\partial {\mathbf{x}}^T\mathbf{x}}{\partial \mathbf{x}}=2\mathbf{x}$$

2 标量对矩阵求导

2.1 定义

设$$f\left(\mathbf{X}\right)=f\left(x_{11},x_{12},\cdots ,x_{1n},x_{21},x_{22},\cdots ,x_{m1,}{x}_{m2},\cdots ,x_{mn}\right)$$
$$\mathbf{X}={\left[\begin{array}{cccc}{x}_{11}& x_{12}& \cdots & x_{1n}\\ x_{21}& x_{22}& \cdots & x_{2n}\\ ⋮& ⋮& \ddots & ⋮\\ x_{m1}& x_{m2}& \cdots & x_{mn}\end{array}\right]}_{m\times n}$$
则：
$$\frac{\partial f}{\partial \mathbf{X}}={\left[\begin{array}{cccc}\frac{\partial f}{\partial x_{11}}& \frac{\partial f}{\partial x_{12}}& \cdots & \frac{\partial f}{\partial x_{1n}}\\ & & & \\ \frac{f}{\partial x_{21}}& \frac{f}{\partial x_{22}}& \cdots & \frac{f}{\partial x_{2n}}\\ & & & \\ ⋮& ⋮& \ddots & ⋮\\ & & & \\ \frac{f}{\partial x_{m1}}& \frac{f}{\partial x_{m2}}& \cdots & \frac{f}{\partial x_{mn}}\end{array}\right]}_{m\times n}$$

2.2 运算法则

2.2.1 线性法则

$$\frac{\partial \left[c_{1}f\left(\mathbf{X}\right)+c_{2}g\left(\mathbf{X}\right)\right]}{\partial \mathbf{X}}=c_1\frac{\partial f\left(\mathbf{X}\right)}{\partial \mathbf{X}}+c_2\frac{\partial g\left(\mathbf{X}\right)}{\partial \mathbf{X}}$$
其中$c_1$与$c_2$为常数。
【证明】
$$\begin{array}{rl}\frac{\partial \left[c_{1}f\left(\mathbf{X}\right)+c_{2}g\left(\mathbf{X}\right)\right]}{\partial \mathbf{X}}& =\left[\begin{array}{cccc}\frac{\partial \left(c_{1}f+c_{2}g\right)}{\partial x_{11}}& \frac{\partial \left(c_{1}f+c_{2}g\right)}{\partial x_{12}}& \cdots & \frac{\partial \left(c_{1}f+c_{2}g\right)}{\partial x_{1n}}\\ & & & \\ \frac{\partial \left(c_{1}f+c_{2}g\right)}{\partial x_{21}}& \frac{\partial \left(c_{1}f+c_{2}g\right)}{\partial x_{22}}& \cdots & \frac{\partial \left(c_{1}f+c_{2}g\right)}{\partial x_{2n}}\\ & & & \\ ⋮& ⋮& \ddots & ⋮\\ & & & \\ \frac{\partial \left(c_{1}f+c_{2}g\right)}{\partial x_{n1}}& \frac{\partial \left(c_{1}f+c_{2}g\right)}{\partial x_{n2}}& \cdots & \frac{\partial \left(c_{1}f+c_{2}g\right)}{\partial x_{nn}}\end{array}\right]\\ & \\ & =\left[\begin{array}{cccc}{c}_1\frac{\partial f}{\partial x_{11}}+c_2\frac{\partial g}{\partial x_{11}}& c_1\frac{\partial f}{\partial x_{12}}+c_2\frac{\partial g}{\partial x_{12}}& \cdots & c_1\frac{\partial f}{\partial x_{1n}}+c_2\frac{\partial g}{\partial x_{1n}}\\ & & & \\ c_1\frac{\partial f}{\partial x_{21}}+c_2\frac{\partial g}{\partial x_{21}}& c_1\frac{\partial f}{\partial x_{22}}+c_2\frac{\partial g}{\partial x_{22}}& \cdots & c_1\frac{\partial f}{\partial x_{2n}}+c_2\frac{\partial g}{\partial x_{2n}}\\ & & & \\ ⋮& ⋮& \ddots & ⋮\\ & & & \\ c_1\frac{\partial f}{\partial x_{n1}}+c_2\frac{\partial g}{\partial x_{n1}}& c_1\frac{\partial f}{\partial x_{n2}}+c_2\frac{\partial g}{\partial x_{n2}}& \cdots & c_1\frac{\partial f}{\partial x_{nn}}+c_2\frac{\partial g}{\partial x_{nn}}\end{array}\right]\\ & \\ & =c_1\left[\begin{array}{cccc}\frac{\partial f}{\partial x_{11}}& \frac{\partial f}{\partial x_{12}}& \cdots & \frac{\partial f}{\partial x_{1n}}\\ & & & \\ \frac{\partial f}{\partial x_{21}}& \frac{\partial f}{\partial x_{22}}& \cdots & \frac{\partial f}{\partial x_{2n}}\\ & & & \\ ⋮& ⋮& \ddots & ⋮\\ & & & \\ \frac{\partial f}{\partial x_{n1}}& \frac{\partial f}{\partial x_{n2}}& \cdots & \frac{\partial f}{\partial x_{nn}}\end{array}\right]+c_2\left[\begin{array}{cccc}\frac{\partial g}{\partial x_{11}}& \frac{\partial g}{\partial x_{12}}& \cdots & \frac{\partial g}{\partial x_{1n}}\\ & & & \\ \frac{\partial g}{\partial x_{21}}& \frac{\partial g}{\partial x_{22}}& \cdots & \frac{\partial g}{\partial x_{2n}}\\ & & & \\ ⋮& ⋮& \ddots & ⋮\\ & & & \\ \frac{\partial g}{\partial x_{n1}}& \frac{\partial g}{\partial x_{n2}}& \cdots & \frac{\partial g}{\partial x_{nn}}\end{array}\right]\\ & \\ & =c_1\frac{\partial f\left(\mathbf{X}\right)}{\partial \mathbf{X}}+c_2\frac{\partial g\left(\mathbf{X}\right)}{\partial \mathbf{X}}\end{array}$$

2.2.2 乘积公式

$$\frac{\partial \left[f\left(\mathbf{X}\right)g\left(\mathbf{X}\right)\right]}{\partial \mathbf{X}}=\frac{\partial f\left(\mathbf{X}\right)}{\partial \mathbf{X}}g\left(\mathbf{X}\right)+f\left(\mathbf{X}\right)\frac{\partial g\left(\mathbf{X}\right)}{\partial \mathbf{X}}$$
【证明】
$$\begin{array}{rl}\frac{\partial \left[f\left(\mathbf{X}\right)g\left(\mathbf{X}\right)\right]}{\partial \mathbf{X}}& =\left[\begin{array}{cccc}\frac{\partial \left(fg\right)}{\partial x_{11}}& \frac{\partial \left(fg\right)}{\partial x_{12}}& \cdots & \frac{\partial \left(fg\right)}{\partial x_{1n}}\\ & & & \\ \frac{\partial \left(fg\right)}{\partial x_{21}}& \frac{\partial \left(fg\right)}{\partial x_{22}}& \cdots & \frac{\partial \left(fg\right)}{\partial x_{2n}}\\ & & & \\ ⋮& ⋮& \ddots & ⋮\\ & & & \\ \frac{\partial \left(fg\right)}{\partial x_{n1}}& \frac{\partial \left(fg\right)}{\partial x_{n2}}& \cdots & \frac{\partial \left(fg\right)}{\partial x_{nn}}\end{array}\right]\\ & \\ & =\left[\begin{array}{cccc}\frac{\partial f}{\partial x_{11}}g+f\frac{\partial g}{\partial x_{11}}& \frac{\partial f}{\partial x_{12}}g+f\frac{\partial g}{\partial x_{12}}& \cdots & \frac{\partial f}{\partial x_{1n}}g+f\frac{\partial g}{\partial x_{1n}}\\ & & & \\ \frac{\partial f}{\partial x_{21}}g+f\frac{\partial g}{\partial x_{21}}& \frac{\partial f}{\partial x_{22}}g+f\frac{\partial g}{\partial x_{22}}& \cdots & \frac{\partial f}{\partial x_{2n}}g+f\frac{\partial g}{\partial x_{2n}}\\ & & & \\ ⋮& ⋮& \ddots & ⋮\\ & & & \\ \frac{\partial f}{\partial x_{n1}}g+f\frac{\partial g}{\partial x_{n1}}& \frac{\partial f}{\partial x_{n2}}g+f\frac{\partial g}{\partial x_{n2}}& \cdots & \frac{\partial f}{\partial x_{nn}}g+f\frac{\partial g}{\partial x_{nn}}\end{array}\right]\\ & \\ & =\left[\begin{array}{cccc}\frac{\partial f}{\partial x_{11}}& \frac{\partial f}{\partial x_{12}}& \cdots & \frac{\partial f}{\partial x_{1n}}\\ & & & \\ \frac{\partial f}{\partial x_{21}}& \frac{\partial f}{\partial x_{22}}& \cdots & \frac{\partial f}{\partial x_{2n}}\\ & & & \\ ⋮& ⋮& \ddots & ⋮\\ & & & \\ \frac{\partial f}{\partial x_{n1}}& \frac{\partial f}{\partial x_{n2}}& \cdots & \frac{\partial f}{\partial x_{nn}}\end{array}\right]g+f\left[\begin{array}{cccc}\frac{\partial g}{\partial x_{11}}& \frac{\partial g}{\partial x_{12}}& \cdots & \frac{\partial g}{\partial x_{1n}}\\ & & & \\ \frac{\partial g}{\partial x_{21}}& \frac{\partial g}{\partial x_{22}}& \cdots & \frac{\partial g}{\partial x_{2n}}\\ & & & \\ ⋮& ⋮& \ddots & ⋮\\ & & & \\ \frac{\partial g}{\partial x_{n1}}& \frac{\partial g}{\partial x_{n2}}& \cdots & \frac{\partial g}{\partial x_{nn}}\end{array}\right]\\ & \\ & =\frac{\partial f\left(\mathbf{X}\right)}{\partial \mathbf{X}}g\left(\mathbf{X}\right)+f\left(\mathbf{X}\right)\frac{\partial g\left(\mathbf{X}\right)}{\partial \mathbf{X}}\end{array}$$

2.2.3 商公式

$$\frac{\partial \left[\frac{f\left(\mathbf{X}\right)}{g\left(\mathbf{X}\right)}\right]}{\partial \mathbf{X}}=\frac{1}{g^2\left(\mathbf{X}\right)}\left[\frac{\partial f\left(\mathbf{X}\right)}{\partial \mathbf{X}}g\left(\mathbf{X}\right)-f\left(\mathbf{X}\right)\frac{\partial g\left(\mathbf{X}\right)}{\partial \mathbf{X}}\right]$$
【证明】
$$\begin{array}{rl}\frac{\partial \left[\frac{f\left(\mathbf{X}\right)}{g\left(\mathbf{X}\right)}\right]}{\partial \mathbf{X}}& =\left[\begin{array}{cccc}\frac{\partial \left(\frac{f}{g}\right)}{\partial x_{11}}& \frac{\partial \left(\frac{f}{g}\right)}{\partial x_{12}}& \cdots & \frac{\partial \left(\frac{f}{g}\right)}{\partial x_{1n}}\\ & & & \\ \frac{\partial \left(\frac{f}{g}\right)}{\partial x_{21}}& \frac{\partial \left(\frac{f}{g}\right)}{\partial x_{22}}& \cdots & \frac{\partial \left(\frac{f}{g}\right)}{\partial x_{2n}}\\ & & & \\ ⋮& ⋮& \ddots & ⋮\\ & & & \\ \frac{\partial \left(\frac{f}{g}\right)}{\partial x_{n1}}& \frac{\partial \left(\frac{f}{g}\right)}{\partial x_{n2}}& \cdots & \frac{\partial \left(\frac{f}{g}\right)}{\partial x_{nn}}\end{array}\right]\\ & \\ & =\left[\begin{array}{cccc}\frac{1}{g^2}\left(\frac{\partial f}{\partial x_{11}}g-f\frac{\partial g}{\partial x_{11}}\right)& \frac{1}{g^2}\left(\frac{\partial f}{\partial x_{12}}g-f\frac{\partial g}{\partial x_{12}}\right)& \cdots & \frac{1}{g^2}\left(\frac{\partial f}{\partial x_{1n}}g-f\frac{\partial g}{\partial x_{1n}}\right)\\ & & & \\ \frac{1}{g^2}\left(\frac{\partial f}{\partial x_{21}}g-f\frac{\partial g}{\partial x_{21}}\right)& \frac{1}{g^2}\left(\frac{\partial f}{\partial x_{22}}g-f\frac{\partial g}{\partial x_{22}}\right)& \cdots & \frac{1}{g^2}\left(\frac{\partial f}{\partial x_{2n}}g-f\frac{\partial g}{\partial x_{2n}}\right)\\ & & & \\ ⋮& ⋮& \ddots & ⋮\\ & & & \\ \frac{1}{g^2}\left(\frac{\partial f}{\partial x_{n1}}g-f\frac{\partial g}{\partial x_{n1}}\right)& \frac{1}{g^2}\left(\frac{\partial f}{\partial x_{n2}}g-f\frac{\partial g}{\partial x_{n2}}\right)& \cdots & \frac{1}{g^2}\left(\frac{\partial f}{\partial x_{nn}}g-f\frac{\partial g}{\partial x_{nn}}\right)\end{array}\right]\\ & \\ & =\frac{1}{g^2}\left(\left[\begin{array}{cccc}\frac{\partial f}{\partial x_{11}}& \frac{\partial f}{\partial x_{12}}& \cdots & \frac{\partial f}{\partial x_{1n}}\\ & & & \\ \frac{\partial f}{\partial x_{21}}& \frac{\partial f}{\partial x_{22}}& \cdots & \frac{\partial f}{\partial x_{2n}}\\ & & & \\ ⋮& ⋮& \ddots & ⋮\\ & & & \\ \frac{\partial f}{\partial x_{n1}}& \frac{\partial f}{\partial x_{n2}}& \cdots & \frac{\partial f}{\partial x_{nn}}\end{array}\right]g-f\left[\begin{array}{cccc}\frac{\partial g}{\partial x_{11}}& \frac{\partial g}{\partial x_{12}}& \cdots & \frac{\partial g}{\partial x_{1n}}\\ & & & \\ \frac{\partial g}{\partial x_{21}}& \frac{\partial g}{\partial x_{22}}& \cdots & \frac{\partial g}{\partial x_{2n}}\\ & & & \\ ⋮& ⋮& \ddots & ⋮\\ & & & \\ \frac{\partial g}{\partial x_{n1}}& \frac{\partial g}{\partial x_{n2}}& \cdots & \frac{\partial g}{\partial x_{nn}}\end{array}\right]\right)\\ & \\ & =\frac{1}{g^2\left(\mathbf{X}\right)}\left[\frac{\partial f\left(\mathbf{X}\right)}{\partial \mathbf{X}}g\left(\mathbf{X}\right)-f\left(\mathbf{X}\right)\frac{\partial g\left(\mathbf{X}\right)}{\partial \mathbf{X}}\right]\end{array}$$
其中
$$g\left(\mathbf{X}\right)\ne 0$$

2.3 示例

【例2】求函数$f\left(\mathbf{A}\right)={\mathbf{x}}^T\mathbf{Ax}$对矩阵$\mathbf{A}$的导数。其中$\mathbf{A}$是对称矩阵，$\mathbf{x}$为三维定常列向量。
【解】
$$\begin{gathered} f\left(\mathbf{A}\right)=\left[\begin{array}{ccc}{x}_1& x_2& x_3\end{array}\right]\left[\begin{array}{ccc}{a}_{11}& a_{12}& a_{13}\\ a_{21}& a_{22}& a_{23}\\ a_{31}& a_{32}& a_{33}\end{array}\right]\left[\begin{array}{c}{x}_1\\ x_2\\ x_3\end{array}\right] \\ =a_{11}\cdot {x_1}^2+a_{22}\cdot {x_2}^2+a_{33}\cdot {x_3}^2+2a_{12}{x}_1\cdot x_2+2a_{13}{x}_1\cdot x_3+2a_{23}{x}_2\cdot x_3 \end{gathered}$$
根据定义，有：
$$\begin{array}{rl}\frac{\partial f\left(\mathbf{A}\right)}{\partial \mathbf{A}}& =\left[\begin{array}{ccc}\frac{\partial f}{\partial a_{11}}& \frac{\partial f}{\partial a_{12}}& \frac{\partial f}{\partial a_{13}}\\ & & \\ \frac{\partial f}{\partial a_{21}}& \frac{\partial f}{\partial a_{22}}& \frac{\partial f}{\partial a_{23}}\\ & & \\ \frac{\partial f}{\partial a_{31}}& \frac{\partial f}{\partial a_{32}}& \frac{\partial f}{\partial a_{33}}\end{array}\right]=\left[\begin{array}{ccc}{x_1}^2& x_1{x}_2& x_1{x}_3\\ & & \\ x_1{x}_2& {x_2}^2& x_2{x}_3\\ & & \\ x_3{x}_1& x_3{x}_2& {x_3}^2\end{array}\right]=\left[\begin{array}{c}{x}_1\\ x_2\\ x_3\end{array}\right]\left[\begin{array}{ccc}{x}_1& x_2& x_3\end{array}\right]\\ & \\ & ={\mathbf{xx}}^T\end{array}$$
即：
$$\frac{\partial {\mathbf{x}}^T\mathbf{Ax}}{\partial \mathbf{A}}={\mathbf{xx}}^T$$

3 向量对标量求导

3.1 定义

设$\mathbf{y}={\left[y_1,y_2,\cdots ,y_m\right]}^T$，在分子布局下，有：
$$\frac{\partial \mathbf{y}}{\partial x}={\left[\frac{\partial y_1}{\partial x},\frac{\partial y_2}{\partial x},\cdots ,\frac{\partial y_m}{\partial x}\right]}^T$$
在分母布局下，有：
$$\frac{\partial \mathbf{y}}{\partial x}=\left[\frac{\partial y_1}{\partial x},\frac{\partial y_2}{\partial x},\cdots ,\frac{\partial y_m}{\partial x}\right]$$

3.2 标量变元的微分

$$\frac{\partial \left(A\pm B\right)}{\partial t}=\frac{\partial A}{\partial t}\pm \frac{\partial B}{\partial t}$$

$$\frac{\partial \left(\lambda A\right)}{\partial t}=\frac{\partial \lambda }{\partial t}A+\lambda \frac{\partial A}{\partial t}$$

$$\frac{\partial \left(a^{T}b\right)}{\partial t}=\frac{\partial a^T}{\partial t}b+a^T\frac{\partial b}{\partial t}$$
$$\frac{\partial \left(AB\right)}{\partial t}=\frac{\partial A}{\partial t}B+A\frac{\partial B}{\partial t}$$

参考文献

[1] 矩阵的求导
[2] 矩阵求导、几种重要的矩阵及常用的矩阵求导公式
[3] 矩阵、向量微分计算