Supervised and Semi-supervised Multi-task Binary Classification

Abstract

In this paper, we interrogate multi-task learning in the background of Gaussian Processes(GP) for constructing different models dealing with the issue of binary classification. At first, we propose a new supervised multi-task classification approach (SMBGC) based on Gaussian processes where kernel parameters for all tasks share a common prior. In recent years great advancement in the field of machine learning domain is being done by exploitation and extraction of information from unlabeled data. Machine learning models require labeled data for training but the amount of labeled data available is quite low since labeling them is expensive. To overcome this problem we came up with a semi-supervised multi-task binary Gaussian process classification (SSMBGC). In this approach, even small amount of labeled data can contribute to our model training and hence they enhance the generalization performance of a model on a learning task with the help of some other related tasks.