Co-EM Support Vector Learning

Abstract

Multi-view algorithms, such as co-training and co-EM, utilize unlabeled datawhen the available attributes can be split into independent and compatiblesubsets. Co-EM outperforms co-training for many problems, but it requires theunderlying learner to estimate class probabilities, and to learn fromprobabilistically labeled data. Therefore, co-EM has so far only been studiedwith naive Bayesian learners. We cast linear classifiers into a probabilisticframework and develop a co-EM version of the Support Vector Machine. Weconduct experiments on text classification problems and compare the family ofsemi-supervised support vector algorithms under different conditions, including violations of the assumptions underlying multi-view learning. Forsome problems, such as course web page classification, we observe the mostaccurate results reported so far.