Abstract

Recent studies have demonstrated the advantages of fusing information from multiple views for vari-ous machine learning applications. However, most existing approaches assumed the shared component common to all views and ignored the private com-ponents of individual views, which thereby restricts the learning performance. In this paper, we pro-pose a new multi-view, low-rank, and sparse ma-trix decomposition scheme to seamlessly integrate diverse yet complementary information stemming from multiple views. Unlike previous approaches, our approach decomposes an input data matrix con-catenated from multiple views as the sum of low-rank, sparse, and noisy parts. Then a unified opti-mization framework is established, where the low-rankness and group-structured sparsity constraints are imposed to simultaneously capture the shared and private components in both instance and view levels. A proven optimization algorithm is devel-oped to solve the optimization, yielding the learned augmented representation which is used as features for classification tasks. Extensive experiments con-ducted on six benchmark image datasets show that our approach enjoys superior performance over the state-of-the-art approaches. 1