「机器学习-李宏毅」:Semi-supervised Learning

这篇文章开篇讲述了什么是Semi-supervised Learning(半监督学习)?

再次,文章具体阐述了四种Semi-supervised Learning,包括Generative Model,Low-density,Smoothness Assumption和Better Representation。

对于Generative Model,文章重点讲述了如何用EM算法来训练模型。

对于Low-density,文章重点讲述了如何让模型进行Self-training,并且在训练中引入Entropy-based Regularization term来尽可能low-density的假设。

对于Smoothness Assumption,文章重点讲述了Graph-based Approach(基于图的方法),并且在训练中引入Smoothness Regularization term来尽可能满足Smoothness Assumption的假设。

对于Better Representation,本篇文章只是简单阐述了其思想,具体介绍见这篇博客。


Your browser is out-of-date!

Update your browser to view this website correctly. Update my browser now

×