矩阵微积分

本文深入探讨矩阵微积分的基本概念和求导规则,包括vector-by-vector、Scalar-by-Vector和scalar-by-scalar的求导形式。介绍了分子布局法和分母布局法,并通过实例详细证明了矩阵对向量、向量对矩阵以及标量乘积的偏导数性质。同时,还给出了求导的一些推论,如∂x∂(a⋅x)的计算。
摘要由CSDN通过智能技术生成
类型 标量 向量 矩阵
标量 ∂ y ∂ x \dfrac{\partial y}{\partial x} xy ∂ y ∂ x \dfrac{\partial \textbf{y}}{\partial x} xy ∂ Y ∂ x \dfrac{\partial Y}{\partial x} xY
向量 ∂ y ∂ x \dfrac{\partial y}{\partial \textbf{x}} xy ∂ y ∂ x \dfrac{\partial \mathbf{y}}{\partial \mathbf{x}} xy
矩阵 ∂ y ∂ X \dfrac{\partial y}{\partial X} Xy

布局方式


对矩阵和向量求导后按两种方式进行组织

  • 分子布局法
  • 分母布局法

f ( x ) = [ f 1 ( x ) f 2 ( x ) ⋮ f n ( x ) ] ,    x ∈ R ,    x ∈ R n ,    X ∈ R m n ,    f i : R n → R ,    f : R n → R n    F : R → R m n \displaystyle \begin{aligned} \mathbf{f(x)} = \begin{bmatrix} f_1(\mathbf{x}) \\ f_2(\mathbf{x}) \\ \vdots \\ f_n(\mathbf{x}) \end{bmatrix}, \; x \in R, \; \mathbf{x} \in R^n, \; X \in R^{mn}, \; f_i:R^n \to R, \; \mathbf{f}:R^n \to R^n \; F:R \to R^{mn} \end{aligned} f(x)=f1(x)f2(x)fn(x),xR,xRn,XRmn,fi:RnR,f:RnRnF:RRmn

形式 分母布局法 分子布局法
∂ f ( x ) ∂ x \dfrac {\partial f(\mathbf{x})}{\partial \mathbf {x} } xf(x) [ ∂ f ( x ) ∂ x 1 ∂ f ( x ) ∂ x 2 ⋮ ∂ f ( x ) ∂ x n ] \begin{bmatrix} \dfrac {\partial f(\mathbf{x})}{\partial x_1} \\[2ex] \dfrac {\partial f(\mathbf{x})}{\partial x_2} \\[2ex] \vdots \\{\dfrac {\partial f(\mathbf{x})}{\partial x_n}}\\ \end{bmatrix} x1f(x)x2f(x)xnf(x) [ ∂ f ( x ) ∂ x 1 ∂ f ( x ) ∂ x 2 ⋯ ∂ f ( x ) ∂ x n ] \begin{bmatrix} \dfrac {\partial f(\mathbf{x})}{\partial x_1} & \dfrac {\partial f(\mathbf{x})}{\partial x_2} & \cdots & \dfrac {\partial f(\mathbf{x})}{\partial x_n} \end{bmatrix} [x1f(x)x2f(x)xnf(x)]
∂ f ( x ) ∂ x \dfrac {\partial \mathbf{f}(x)}{\partial x} xf(x) [ ∂ f 1 ( x ) ∂ x ∂ f 2 ( x ) ∂ x ⋯ ∂ f m ( x ) ∂ x ] \begin{bmatrix} \dfrac {\partial f_1(x)}{\partial x} &\dfrac {\partial f_2(x)}{\partial x} &\cdots &\dfrac {\partial f_m(x)}{\partial x} \end{bmatrix} [xf1(x)xf2(x)xfm(x)] [ ∂ f 1 ( x ) ∂ x ∂ f 2 ( x ) ∂ x ⋮ ∂ f m ( x ) ∂ x ] \begin{bmatrix} \dfrac {\partial f_1(x)}{\partial x} \\[2ex] \dfrac {\partial f_2(x)}{\partial x} \\[2ex] \vdots \\{\dfrac {\partial f_m(x)}{\partial x}}\\ \end{bmatrix} xf1(x)xf2(x)xfm(x)
∂ f ( x ) ∂ x \dfrac {\partial \mathbf{f(x)}}{\partial \mathbf{x}} xf(x) [ ∂ f 1 ( x ) ∂ x ∂ f 2 ( x ) ∂ x ⋯ ∂ f m ( x ) ∂ x ] [ ∂ f ( x ) ∂ x 1 ∂ f ( x ) ∂ x 2 ⋮ ∂ f ( x ) ∂ x n ] [ ∂ f 1 ( x ) ∂ x 1 ∂ f 2 ( x ) ∂ x 1 ⋯ ∂ f m ( x ) ∂ x 1 ∂ f 1 ( x ) ∂ x 2 ∂ f 2 ( x ) ∂ x 2 ⋯ ∂ f m ( x ) ∂ x 2 ⋮ ⋮ ⋱ ⋮ ∂ f 1 ( x ) ∂ x n ∂ f 2 ( x ) ∂ x n ⋯ ∂ f m ( x ) ∂ x n ] \begin{bmatrix} \dfrac {\partial f_1(\mathbf{x})}{\partial \mathbf{x}} &\dfrac {\partial f_2(\mathbf{x})}{\partial \mathbf{x}} &\cdots &\dfrac {\partial f_m(\mathbf{x})}{\partial \mathbf{x}} \end{bmatrix} \\[2ex] \begin{bmatrix} \dfrac {\partial \mathbf{f(x)}}{\partial x_1} \\[2ex] \dfrac {\partial \mathbf{f(x)}}{\partial x_2} \\[2ex] \vdots \\{\dfrac {\partial \mathbf{f(x)}}{\partial x_n}}\\ \end{bmatrix} \\[2ex] \begin{bmatrix} \dfrac {\partial f_1(\mathbf{x})}{\partial x_1} &\dfrac {\partial f_2(\mathbf{x})}{\partial x_1} &\cdots &\dfrac {\partial f_m(\mathbf{x})}{\partial x_1} \\[2ex] \dfrac {\partial f_1(\mathbf{x})}{\partial x_2} &\dfrac {\partial f_2(\mathbf{x})}{\partial x_2} &\cdots &\dfrac {\partial f_m(\mathbf{x})}{\partial x_2} \\[2ex] \vdots &\vdots &\ddots &\vdots \\[2ex] \dfrac {\partial f_1(\mathbf{x})}{\partial x_n} &\dfrac {\partial f_2(\mathbf{x})}{\partial x_n} &\cdots &\dfrac {\partial f_m(\mathbf{x})}{\partial x_n} \end{bmatrix} [xf1(x)xf2(x)xfm(x)]x1f(x)x2f(x)xnf(x)x1f1(x)x2f1(x)xnf1(x)x1f2(x)x2f2(x)xnf2(x)x1fm(x)x2fm(x)xnfm(x) [ ∂ f 1 ( x ) ∂ x 1 ∂ f 2 ( x ) ∂ x 2 ⋯ ∂ f m ( x ) ∂ x n ] [ ∂ f 1 ( x ) ∂ x ∂ f 2 ( x ) ∂ x ⋮ ∂ f m ( x ) ∂ x ] [ ∂ f 1 ( x ) ∂ x 1 ∂ f 1 ( x ) ∂ x 2 ⋯ ∂ f 1 ( x ) ∂ x n ∂ f 2 ( x ) ∂ x 1 ∂ f 2 ( x ) ∂ x 2 ⋯ ∂ f 2 ( x ) ∂ x n ⋮ ⋮ ⋱ ⋮ ∂ f m ( x ) ∂ x 1 ∂ f m ( x ) ∂ x 2 ⋯ ∂ f m ( x ) ∂ x n ] \begin{bmatrix} \dfrac {\partial f_1(\mathbf{x})}{\partial x_1} &\dfrac {\partial f_2(\mathbf{x})}{\partial x_2} &\cdots &\dfrac {\partial f_m(\mathbf{x})}{\partial x_n} \end{bmatrix} \\[2ex] \begin{bmatrix} \dfrac {\partial f_1(\mathbf{x})}{\partial \mathbf{x}} \\[2ex] \dfrac {\partial f_2(\mathbf{x})}{\partial \mathbf{x}} \\[2ex] \vdots \\{\dfrac {\partial f_m(\mathbf{x})}{\partial \mathbf{x}}}\\ \end{bmatrix} \\[2ex] \begin{bmatrix} \dfrac {\partial f_1(\mathbf{x})}{\partial x_1} &\dfrac {\partial f_1(\mathbf{x})}{\partial x_2} &\cdots &\dfrac {\partial f_1(\mathbf{x})}{\partial x_n} \\[2ex] \dfrac {\partial f_2(\mathbf{x})}{\partial x_1} &\dfrac {\partial f_2(\mathbf{x})}{\partial x_2} &\cdots &\dfrac {\partial f_2(\mathbf{x})}{\partial x_n} \\[2ex] \vdots &\vdots &\ddots &\vdots \\[2ex] \dfrac {\partial f_m(\mathbf{x})}{\partial x_1} &\dfrac {\partial f_m(\mathbf{x})}{\partial x_2} &\cdots &\dfrac {\partial f_m(\mathbf{x})}{\partial x_n} \end{bmatrix} [x1f1(x)x2f2(x)xnfm(x)]xf1(x)xf2(x)xfm(x)x1f1(x)x1f2(x)x1fm(x)x2f1(x)x2f2(x)x2fm(x)xnf1(x)xnf2(x)xnfm(x)
∂ f ( X ) ∂ X \dfrac {\partial f(X)}{\partial X}
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值