Margin-based discriminant embedding guided sparse matrix regression for image supervised feature selection

Xiuhong Chen, Yun Lu, Jun Zhang, Xingyu Zhu

Abstract: Highlights•Our model preserves the spatial information of elements in the original image data.•The margin of each matrix data is defined by using left/right regression matrices.•Maximizing the average margin can obtain the non-linear discriminant embedding.•Discriminative features are selected by the row sparsity of transformation matrix.•An iterative optimization algorithm is designed to solve the proposed model.•The effectiveness and superiority of our method are verified on several datasets.