TLD

PN-learning:Bootstrapping Binary Classifiers by Structural Constraints, Zdenek Kalal,CVPR2010

PN学习可以定义为一下的过程： 

（1）准备一个数量较少的训练样本集合和一个数量很大的测试样本集合。 

（2）利用训练样本训练一个初始分类器。同时，利用训练样本对(先验)约束条件进行相应的调整。 

（3）利用分类器对测试样本赋予标签，并找出分类器赋予的标签同约束条件相矛盾的那些样本； 

（4）将上述相矛盾的样本重新赋予标签，将其加入训练样本，重新训练分类器； 

反复迭代上述过程，直到满足某个约束条件

Exploiting unlabeled data

Expectation-Maximization (EM) is an algorithm for finding parameters of probabilistic models.EM is sometimes interpreted as a “soft” version of self-learning.

Self-learning is probably the oldest approach for semisupervised learning.