Discriminative Feature Selection via A Structured Sparse Subspace Learning Module

In this paper, we first propose a novel Structured Sparse Subspace Learning ((SL)-L-3) module to address the long-standing subspace sparsity issue. Elicited by proposed module, we design a new discriminative feature selection method, named Subspace Sparsity Discriminant Feature Selection ((SDFS)-D-2) which enables the following new functionalities: 1) Proposed (SDFS)-D-2 method directly joints trace ratio objective and structured sparse subspace constraint via l(2,0) -norm to learn a rowsparsity subspace, which improves the discriminability of model and overcomes the parametertuning trouble with comparison to the methods used l(2,0)-norm regularization; 2) An alternative iterative optimization algorithm based on the proposed (SL)-L-3 module is presented to explicitly solve the proposed problem with a closed-form solution and strict convergence proof. To our best knowledge, such objective function and solver are first proposed in this paper, which provides a new though for the development of feature selection methods. Extensive experiments conducted on several high-dimensional datasets demonstrate the discriminability of selected features via (SDFS)-D-2 with comparison to several related SOTA feature selection methods. Source matlab code: haps:// github. com/StevenWangNPU/L20-FS.