Multi-view subspace learning via bidirectional sparsity

Ruidong Fan, Tingjin Luo, Wenzhang Zhuge, Sheng Qiang, Chenping Hou

Abstract: Highlights•By using the shared subspace’s bidirectional sparsity, we propose a novel approach which can find an effectivesubspace dimension and deal with outliers simultaneously.•We design an effective algorithm to solve the non-convex model and give its convergence analysis.•Compared with other multi-view subspace learning methods, the extensive experimental results on real-worlddatasets present the effectiveness of our model.