$L_{\alpha}$-Regularization-Based Sparse Semi-Supervised Learning for Data with Complex Distributions
Qi Zhang, Tianguang Chu · 2019 IEEE 8th Data Driven Control and Learning Systems Conference (DDCLS) · 2019
We consider in this paper two kinds of complex data, i.e., multimodal and mixmodal data, and aim to develop data-driven learning models to exploit the intrinsic characteristics of the data. An lα-regularization-based sparse semi-supervised graph embedding (SSGE) model for feature extraction upon multimodal and mixmodal data is presented by incorporating the hierarchical locality with non-convex sparsity, facilitating better interpretation of the projections. Semi-supervised projection learning and lα-regularization-based sparse subspace learning are implemented successively, with feasible solving algorithms presented. Experiments for face recognition verifies the feasibility and effectiveness of the proposed SSGE model in both multimodal and mixmodal situations.