Semi-supervised Regression with Order Preferences

Abstract

Following a discussion on the general form of regularization for semi-supervised learning, we propose a semi-supervised regression algorithm. It is based on the assumption that we have certain order preferences on unlabeled data (e.g., point X1 has a larger target value than x2). Semi-supervised learning consists of enforcing the order preferences as regularization in a risk minimization framework. The optimization problem can be effectivley solved by a linear program. Experiments show that the proposed semi-supervised regression outperforms standrad regression.