Structured discriminative concept factorization for data representation

Abstract

Concept factorization has attracted much attention in the past few years. To consider the manifold structure embedded in data, the graph regularizer is incorporated into model of concept factorization. However, a single graph cannot effectively model the intrinsic structure information of data. To solve this problem, a novel method, called Structured Discriminative Concept Factorization (SDCF), is proposed to explore the intrinsic structure information of data. Specifically, the proposed SDCF method incorporates both the local affinity and the distant repulsion constraints into the model of CF. Moreover, an efficient optimization scheme based on multiple update algorithm for the proposed SDCF method is developed. Experimental results on benchmark datasets have validated the effectiveness of the proposed method.