Capped l2,1-Norm Regularization on Kernel Regression for Robust Semi-supervised Learning

Abstract

Graph-based semi-supervised learning has become one of the most important research areas in machine learning and artificial intelligence community. In this paper, we propose a Capped l2,1 -Norm Regularization model for graph based semi-supervised learning (SSL). The new proposed model is aims to fully train the classification function by utilizing all the data points as well as handle the out-of-sample extension for new-coming data points. In addition, in order to enhance the robustness to the outliers, we leverage the capped l2,1 -norm as the loss function for the classification model. The capped l2,1 norm can suppress the bias of outliers that are far away from the normal data distribution. Simulation results show that the proposed method can achieve better performance compared with other state-of-the-art graph based SSL methods.