"
1. 简介
生成式模型(generative model)会对xxx和yyy的联合分布p(x,y)p(x,y)p(x,y)进行建模,然后通过贝叶斯公式来求得p(y|x)p(y|x)p(y|x), 最后选取使得p(y|x)p(y|x)p(y|x)最大的yiyiy_i. //代码效果参考:https://v.youku.com/v_show/id_XNjQwNjYyODc4NA==.html
具体地, y?=argmaxyip(yi|x)=argmaxyip(x|yi)p(yi)p(x)=argmaxyip(x|yi)p(yi)=argmaxyip(x,yi)y?=argmaxyip(yi|x)=argmaxyip(x|yi)p(yi)p(x)=argmaxyip(x|yi)p(yi)=argmaxyip(x,yi)y{*}=arg \max{y_i}p(yi|x)=arg \max{y_i}\frac{p(x|y_i)p(yi)}{p(x)}=arg \max{y_i}p(x|y_i)p(yi)=arg \max{y_i}p(x,y_i).//代码效果参考:https://v.youku.com/v_show/id_XNjQwMDE5ODQ3Ng==.html 判别式模型(discriminative model)则会直接对"