三道试题 第一题
![三道试题](https://i-blog.csdnimg.cn/blog_migrate/820fd7ce71eb37b8d2e6966b550da067.png)
不是序
从15年开始真正接触机器学习,也尝试着看了一些书,比如《机器学习》(周志华)、《统计机器学习》(李航)、《机器学习实战》等等;一直对于相关数学应该掌握到什么程度心存疑虑,有的数学相对艰深,推导和证明比较花时间,但是直接略过之后,发现看后续的内容会雪崩式的夹生饭;直到学完了Stanford CS229《Machine Learning》,逼着自己在规定时间内做完所有的习题和编程作业,觉得比较有收获,所谓吾尝终日而思矣,不如须臾之所学也。
机器学习的重要的数学基础之一是线性代数和多变量微积分(Linear Algebra & Multivariable Calculus)。对于这门学科要掌握到什么程度?请弄懂三道试题即可。
第一题:Gradients and Hessians
题目
Recall that a matrix A ∈ R n × n A\in{R}^{n\times n} A∈Rn×n is symmetric if A T = A A^T=A AT=A, that is, A i j = A j i {A}_{ij}={A}_{ji} Aij=Aji for all i, j. Also
recall the gradient ▽ f ( x ) \triangledown f(x) ▽f(x) of a function f : R n → n f:{R}^{n} \rightarrow n f:Rn→n which is the n-vector of partial derivatives. ▽ f ( x ) = [ ∂ ∂ x 1 f ( x ) ⋮ ∂ ∂ x n f ( x ) ] w h e r e x = [ x 1 ⋮ x n ] \triangledown f(x) = \begin{bmatrix} \frac { \partial }{ \partial { x }_{ 1 } }f(x)\\ \vdots\\ \frac { \partial }{ \partial { x }_{ n }}f(x) \end{bmatrix} \quad where \quad x = \begin{bmatrix} {x}_{1} \\ \vdots \\ {x}_{n} \end{bmatrix} ▽f(x)=⎣⎢⎡∂x1∂f(x)⋮∂xn∂f(x)⎦⎥⎤wherex=⎣⎢⎡x1⋮xn⎦⎥⎤
The hessian ▽ 2 f ( x ) {\triangledown}^{2}f(x) ▽2f(x) of a function f : R n → R f:{R}^{n} \rightarrow R f:Rn→R is the n × n n \times n n×n symmetric matrix of twice partial derivatives,
[ ∂ 2 ∂ x 1 2 f ( x ) ∂ 2 ∂ x 1 x 2 f ( x ) ⋯ ∂ 2 ∂ x 1 ∂ x n f ( x ) ∂ 2 ∂ x 2 x 1 f ( x ) ∂ 2 ∂ x 2 2 f ( x ) ⋯ ∂ 2 ∂ x 2 ∂ x n f ( x ) ⋮ ⋮ ⋱ ⋮ ∂ 2 ∂ x n x 1 f ( x ) ∂ 2 ∂ x n x 2 f ( x ) ⋯ ∂ 2 ∂ x n 2 f ( x ) ] \begin{bmatrix} \frac{
{\partial}^{2}}{\partial {x}_{1}^{2}}f(x) & \frac{
{\partial}^{2}}{\partial {x}_{1} {x}_{2}}f(x) & \cdots & \frac{
{\partial}^{2}}{\partial {x}_{1} \partial {x}_{n}}f(x) \\ \frac{
{\partial}^{2}}{\partial {x}_{2}{x}_{1}}f(x) & \frac{
{\partial}^{2}}{\partial {x}_{2}^{2}}f(x) & \cdots & \frac{
{\partial}^{2}}{\partial {x}_{2} \partial {x}_{n}}f(x) \\ \vdots & \vdots & \ddots &\vdots \\ \frac{
{\partial}^{2}}{\partial {x}_{n}{x}_{1}}f(x) & \frac{
{\partial}^{2}}{\partial {x}_{n}{x}_{2}}f(x) & \cdots & \frac{
{\partial}^{2}}{\partial {x}_{n}^{2}}f(x) \end{bmatrix} ⎣⎢⎢⎢⎢⎢⎡∂x12∂2f(x)∂x2x1∂2f(x)⋮∂xnx1∂2f(x)∂x1x2∂2f(x)∂x22∂2f(x)⋮∂xnx2∂2f(x)⋯⋯⋱⋯∂x1∂xn∂2f(x)∂x2∂xn∂2f(x)⋮∂xn2∂2f(x)⎦⎥⎥⎥⎥⎥⎤
第一问
Let f ( x ) = 1 2 x T A x + b T x f(x)=\frac{1}{2}{x}^{T}Ax + {b}^{T}x f(x)=21xTAx+bTx where A is a symmetric matrix and b ∈ R n b \in {R}^{n} b∈Rn is a vector. What is ▽ f ( x ) \triangledown f(x) ▽f(x)?
第一答
设 A = [ R 1 R 2 ⋯ R n ] A = \begin{bmatrix} {R}_{1} \\ {R}_{2} \\ \cdots \\ {R}_{n} \end{bmatrix} A=⎣⎢⎢⎡R1R2⋯Rn⎦⎥⎥⎤ 其中: R i = [ A i 1 A i 2 ⋯ A i n ] {R}_{i} = \begin{bmatrix} {A}_{i1} \quad {A}_{i2} \quad \cdots \quad {A}_{in} \end{bmatrix} Ri=[Ai1Ai2⋯Ain]
设 f 1 ( x ) = x T A x = x T ( A x ) ⋯ 矩 阵 乘 法 结 合 律 = x T ( [ R 1 R 2 ⋯ R n ] x ) = x T [ R 1 x R 2 x ⋯ R n x ] = x T [ R 1 x R 2 x ⋯ R n x ] ⋯ R i 为 行 向 量 ; x 为 列 向 量 = [ x 1 x 2 ⋯ x n ] [ R 1 x R 2 x ⋮ R n x ] = ∑ i = 1 n x i R i x = ∑ i = 1