Skip to main content

雅可比(Jacobi)矩阵、海塞(Hessan)矩阵

· 3 min read
PuQing
AI, CVer, Pythoner, Half-stack Developer

一、雅可比(Jacobi)矩阵

对于 nn 个变元的 mm 个函数

y1=f1(x1,x2,,xn),y2=f2(x1,x2,,xn),ym=fm(x1,x2,,xn),}\left.\begin{array}{l} y_{1}=f_{1}\left(x_{1}, x_{2}, \cdots, x_{n}\right), \\ y_{2}=f_{2}\left(x_{1}, x_{2}, \cdots, x_{n}\right), \\ \cdots \cdots \cdots \cdots \cdots \cdots \cdots \\ y_{m}=f_{m}\left(x_{1}, x_{2}, \cdots, x_{n}\right), \end{array}\right\}

它定于于某一 nn 维区域 D\mathcal{D} 中,并且在这一区域中有关于一切变元的连续偏导数

则定义 Jacobi\mathrm{Jacobi} 矩阵为:

J=[fx1fxn]=[f1x1f1xnfmx1fmxn]\mathbf{J}=\begin{bmatrix} \frac{\partial \mathbf{f}}{\partial x_{1}} & \cdots & \frac{\partial \mathbf{f}}{\partial x_{n}} \end{bmatrix}=\begin{bmatrix} \frac{\partial f_{1}}{\partial x_{1}} & \cdots & \frac{\partial f_{1}}{\partial x_{n}} \\ \vdots & \ddots & \vdots \\ \frac{\partial f_{m}}{\partial x_{1}} & \cdots & \frac{\partial f_{m}}{\partial x_{n}} \end{bmatrix}

分量可以表示为

Jij=fjxj\mathbf{J}_{ij}=\frac{\partial f_j}{\partial x_j}

或者写为:Jf(x1,,xn)\mathbf{J}_{\mathbf{f}}\left(x_{1}, \ldots, x_{n}\right) 或者 (f1,,fm)(x1,,xn)\displaystyle\frac{\partial\left(f_{1}, \ldots, f_{m}\right)}{\partial\left(x_{1}, \ldots, x_{n}\right)}

二、海塞(Hessan)矩阵

对于函数 f(x)f(x),其中的 x=(x1;x2;x3,;xn)x=\left(x_{1} ; x_{2} ; x_{3}, \ldots ; x_{n}\right),其定义的 Hessan\mathrm{Hessan} 矩阵为:

H=(fx1x1fx1x2fx1xnfx2x1fx2x2fx2xnfxnx1fxnx2fxnxn)H=\left(\begin{array}{cccc} \frac{\partial f}{\partial x_{1} \partial x_{1}} & \frac{\partial f}{\partial x_{1} \partial x_{2}} & \cdots & \frac{\partial f}{\partial x_{1} \partial x_{n}} \\ \frac{\partial f}{\partial x_{2} \partial x_{1}} & \frac{\partial f}{\partial x_{2} \partial x_{2}} & \cdots & \frac{\partial f}{\partial x_{2} \partial x_{n}} \\ \vdots & \vdots & \ddots & \vdots \\ \frac{\partial f}{\partial x_{n} \partial x_{1}} & \frac{\partial f}{\partial x_{n} \partial x_{2}} & \cdots & \frac{\partial f}{\partial x_{n} \partial x_{n}} \end{array}\right)
Hessan matrix\mathrm{Hessan\ matrix}Jacobi matrix\mathrm{Jacobi\ matrix} 的关系
info
Hf=J(f)H_f = J(\nabla f^\top)

即对一个求了偏导得到的一阶偏导函数向量求 Jacobi\mathrm{Jacobi} 矩阵就是 Hessan\mathrm{Hessan} 矩阵

What is \nabla
对称性

如果函数 ff 在区域 D\mathcal{D} 内二阶连续可导,那么 ffHessan\mathrm{Hessan} 矩阵在 D\mathcal{D} 内为 对称矩阵(This page is not published)

证明:

如果函数 ff 连续,则二阶偏导数的求导顺序没有区别,即:

x(fy)=y(fx)\frac{\partial}{\partial x}\left(\frac{\partial f}{\partial y}\right)=\frac{\partial}{\partial y}\left(\frac{\partial f}{\partial x}\right)

则对于 Hessan\mathrm{Hessan} 矩阵 H(f)H(f),有 Hi,j(f)=Hj,i(f)H_{i,j}(f)=H_{j,i}(f),所以 H(f)H(f) 为对称矩阵

相关资料