Time: 2019-06-24 10：00-11：30

Venue: Room 201 Lecture Hall

Abstract:

Cross-view classification that means to classify samples from heterogeneous views is a significant yet challenging problem in computer vision.

An effective solution to this problem is the multi-view subspace learning (MvSL), which intends to find a common subspace for multi-view data. Despite promising results obtained on some applications, the performance of existing methods deteriorates dramatically when the multi-view data is sampled from nonlinear manifolds. To circumvent this drawback, we propose Multi-view Hybrid Embedding (MvHE) and Multi-view Common Component Discriminant Analysis (MvCCDA) algorithm to handle view discrepancy, discriminancy and nonlinearity simultaneously. Extensive experiments demonstrate the overwhelming advantages against the state-of-the-art MvSL based approaches in terms of classification accuracy.

Low-rank Multi-view Subspace Learning (LMvSL) has also shown great potential in cross-view classification in recent years. Despite their empirical success, existing LMvSL based methods fail to address view discrepancy and discriminancy simultaneously and are incapable of handling complicated noise in practice. To alleviate such limitation, we propose Modal Regression based Structured Low-rank Matrix Recovery (MR-SLMR), a unique method of effectively removing view discrepancy and improving discriminancy through the recovery of structured low-rank matrix. Moreover, modal regression incorporated into our model ensures that MR-SLMR is robust to complicated noise. Experimental results demonstrate the superiority of MR-SLMR and its robustness to complicated noise.