Semi-supervised learning using a constrained labeling LDA model

In problems where labeled data is scarce, semi-supervised learning (SSL) techniques are an attractive framework that can exploit both labeled and unlabeled data. In this paper, we introduce an alternate version of a semi-supervised algorithm, the generative approach linear discriminant analysis (LDA) approach is used. We provide empirical results of using this method on synthetic and several benchmark data sets, then we analyze the behavior of the method.