zoukankan      html  css  js  c++  java
  • 导论(习题)

    1 导论

    1.1

    [ abla_mathbf{w}E(mathbf{w})=sum_{n=1}^N{y(x_n,mathbf{w})-t_n}phi(x_n)=0 ]

    [sum_{n=1}^Nphi^ ext{T}(x_n)mathbf{w}phi(x_n)=sum_{n=1}^Nt_nphi(x_n) ]

    [mathbf{w}sum_{n=1}^Nphi(x_n)phi^ ext{T}(x_n) = sum_{n=1}^Nt_nphi(x_n) ]

    1.2

    1.29

    由于(ln(cdot))是凹函数,利用琴生不等式,

    [egin{aligned} ext{H}[x]&=-sum_{i=1}^Mp_iln p_i\&=sum_{i=1}^Mp_iln frac{1}{p_i}\&leqlnleft(sum_{i=1}^Mp_icdot frac{1}{p_i} ight)\&=ln M.end{aligned} ]

    1.30

    [egin{aligned} ext{KL}(pVert q)&=-int p(x)lnfrac{q(x)}{p(x)}{ m{d}}x\&=-int p(x)lnleft(frac{sigma}{s}expleft(-frac{1}{2}left[frac{(x-m)^2}{s^2}-frac{(x-mu)^2}{sigma^2} ight] ight) ight){ m{d}}x\&=-int p(x)left[lnfrac{sigma}{s}-frac{1}{2}left(frac{(x-m)^2}{s^2}-frac{(x-mu)^2}{sigma^2} ight) ight]{ m{d}}x \&=lnfrac{s}{sigma}-frac{1}{2}intleft[left(frac{1}{sigma^2}-frac{1}{s^2} ight)x^2+2left(frac{m}{s^2}-frac{mu}{sigma^2} ight)x+left(frac{mu^2}{sigma^2}-frac{m^2}{s^2} ight) ight]p(x){ m{d}}x \&=lnfrac{s}{sigma}+frac{1}{2}+frac{sigma^2+(mu-m)^2}{2s^2}. end{aligned}]

  • 相关阅读:
    0317复利计算的回顾与总结
    0518 Scrum 项目 5.0
    0517 Scrum 项目4.0
    0512 Scrum 项目3.0
    实验三 进程调度模拟程序
    0505 Scrum 项目1.0
    0502团队项目 SCRUM团队成立
    0428 团队项目2.0
    0422团队项目
    实验二 作业调度模拟程序
  • 原文地址:https://www.cnblogs.com/astoninfer/p/9253370.html
Copyright © 2011-2022 走看看