Zhao Zhang

TL;DR: A novel graph-regularized matrix factorization model is developed to preserve the local geometric similarities of the learned common representations from different views and the semantic consistency constraint is introduced to stimulate these view-specific representations toward a unified discriminative representation.

TL;DR: Two nuclear- and L2,1-norm regularized 2D neighborhood preserving projection methods for extracting representative 2D image features and imposing the low-rank or sparse constraints on projections at the same time can outperform related state-of-the-arts in a variety of simulation settings.

TL;DR: The classification approach of the ADDL model is very efficient, because it can avoid the extra time-consuming sparse reconstruction process with trained dictionary for each new test data as most existing DL algorithms.

TL;DR: An extensive empirical study on baseline encoder-decoder models in terms of different encoder backbones, loss functions, training batch sizes, and attention structures is presented, and new baseline models that can outperform state-of-the-art performance were discovered.

TL;DR: This work proposes a transductive low-rank and sparse principal feature coding (LSPFC) formulation that decomposes given data into a component part that encodes low- rank sparse principal features and a noise-fitting error part, and presents an inductive LSPFC (I-L SPFC), which incorporates embedded low- Rank and sparse Principal features by a projection into one problem for direct minimization.