用正则的方法在正样本和无标签样本上训练二分类器

Training Bi-classifier on Positive and Unlabeled Sample by Regular Method

正样本-无标签样本学习(Positive-Unlabelled Learning,简称PU Learning)。相对于标准的正样本-负样本学习(Positive-NegativeLearning,简称PNLearning),其是一种仅使用少量正样本和大量无标签样本训练二类分类器的方法。以往工作通常使用一种无偏估计的方法,使用正样本和无标签样本估计PNLearning损失,并建立深度神经网络模型。但是,这种估计方法依赖正样本的数量,当正样本数量较少时,学习算法会发生严重的过拟合。

Positive sample-unlabeled sample(PU Learning).Compared with the standard positive-negative learning(PN Learning),it is a method of training class II classifiers using only a small number of positive samples and a large number of unlabeled samples.Previous work usually used an unbiased estimation method,using positive and unlabeled samples to estimate PN Learning loss,and built a deep neural network model.However,this estimation method depends on the number of positive samples.When the number of positive samples is small,the learning algorithm will be seriously over-fitting.