Multi-label feature selection via adaptive dual-graph optimization

Zhenzhen Sun, Hao Xie, Jinghua Liu, Yuanlong Yu

Abstract: Highlights•A dual-graph regularization is designed to capture local label correlations.•A non-convex sparse metric l2,1−2<math><msub is="true"><mrow is="true"><mi is="true">l</mi></mrow><mrow is="true"><mn is="true">2</mn><mo is="true">,</mo><mn is="true">1</mn><mo is="true">−</mo><mn is="true">2</mn></mrow></msub></math>-norm is employed to select low-redundancy feature subset.•An adaptive spectral graph is implemented during the feature selection process.•An alternate iterative updating rule for solving the optimization problem is designed.