Semi-supervised Discriminative Structured Prediction

Abstract

Abstract : The proposed research develops cutting-edge machine learning techniques to improve the performance of a wide spectrum of robust and intelligent classification tasks. Structured prediction, one of major challenges in statistical machine learning, is a classification or regression problem with non-iid data where the prediction variables are typically interdependent in complex ways with dependencies encoded in a graphical model to capture the sequential, spatial, relational or recursive structure of output variables. Semi-supervised learning, another example of the four major challenges in statistical machine learning, is a technique which makes use of both unlabeled and labeled data for training typically a small amount of labeled data with a large amount of unlabeled data. Traditinal approaches optimize surrogate functions of performance measures for structured prediction. In this project, we propose to design novel machine learning algorithms that directly optimize performance measures for classification and ranking problems.