szcf-weiya / esl-cn Goto Github PK

where: section 7.4

how to show?

details: section 5.2
Is there an example to support the claim that "terms like this are typically included with other terms in the model."?

结合http://ai.stanford.edu/~paskin/gm-short-course/， 完善有关 junction tree algorithm的问题。

Refer to 15.2 Definition of Random Forests for more details.

如图

TODO: section 8.7.1

When the resolution ratio becomes smaller, then the size of math formula would exceed the max width of the device, e.g. mobile phone, and this will let this web page ugly.

e.g.

to do

TODO: chap9.1 about IRLS algorithm.

In section 6.9, it used flops to measure the computational cost.

6.8 Mixture Models for Density Estimation and Classification
& 6.9 Computational Considerations

In section 11.7, it claims that "The training set error for all of the networks was 0%, since in all cases there are more parameters than training observations." (page 406)

Why??

to do

Todo

最大值处为什么是p(x)？

页面上无法显示该条评论，但是在disqus后台可以，之前陆陆续续尝试了disqus中的migrate tools，都没有解决

to do

to do

part of 3.8 More on the Lasso and Related Path Algorithms
& 3.9 Computational Considerations

when visiting the home page

The materials from Prof. Liu have something with BF, and section 7.7 also discusses BF. Try to associate them.

In fig. 7.2 of section 7.3, what is the meaning of the red line? the closest fit?
And what is the difference between "closest fit" and "closet fit in population"?
Is estimation bias of the models on the red line zero?

details: section 5.2.3

进度

更新于2017.03.13

未完成章节

• chap03: 3.7-3.9;
• chap05: 5.6-5.9;
• chap06: 6.7-6.9;
• chap08: 8.8-8.9;
• chap10: 10.7-10.14;
• chap11: 11.7-11.10;
• chap12: 12.3-12.7;
• chap13: 13.2-13.5;
• chap14: 14.4-14.10;
• chap15: 15.2-15.4;
• chap16: 16.1-16.3;
• chap18: 18.1-18.8;

TODO: 5.5 Automatic Selection of the Smoothing Parameters

details: section 15.3

16.3节提到的post-processing boosting和lasso post processor

用R语言实现chap18.1的模拟

18.6节中提到的cox‘s proportional hazards regression, 有时间学习一下。