Semi-Supervised Similarity Preserving Co-Selection

Abstract

Semi-supervised learning is the required paradigm when data are partially labeled. It is more adapted for large domain applications when labels are hardly and costly to obtain. In addition, when data are large, feature selection and instance selection are two important dual operations for removing irrelevant information. To address theses challenges together, we propose a unified framework, called sCOs, for semi-supervised co-selection of features and instances, simultaneously. In particular, we propose a novel cost function based on l2, 1-norm regularization and similarity preserving selection of both features and instances. Experimental results on some known benchmark datasets are provided for validating sCOs and comparing it with some representative methods in the state-of-the art.