The performance of semi-supervised self-training classifier largely depends on the quality of pseudo labels.This paper proposes a robust semi-supervised classification method for logistic regression with l2 regularization,using a random forest training residual model and robust Mahalanobis distance to improve the quality of pseudo-labels.A large number of experiments have been carried out to evaluate the algorithm.
关键词
半监督分类/自训练分类/稳健马氏距离/随机森林/残差模型
Key words
semi-supervised classification/self-training/robust Mahalanobis distance/random forest/residual model
维普
万方数据