Data-dependent kernel sparsity preserving projection and its application for semi-supervised classification

作者：Zhang, Ao; Gao, Xianwen^*

来源：Multimedia Tools and Applications, 2018, 77(18): 24459-24475.

DOI：10.1007/s11042-018-5707-0

• Sparsity preserving projection
• Data-dependent kernel
• Semi-supervised learning
• Dimensionality reduction

Dimensionality reduction methods (DR) have been commonly used as a principled way to understand the high-dimensional data. In this paper, a novel semi-supervised nonlinear method called semi-supervised data-dependent kernel sparsity preserving projection (SDKSPP) is proposed for dimensionality reduction. To achieve performance improvements, SDKSPP adopts a data-dependent kernel (DK) instead of a standard kernel. The coefficients in DK are optimized with labeled samples by using the Fisher criterion. Then the labeled and unlabeled samples are mapped into a high dimensional space by DK. The sparse reconstructive relationship among the whole samples is calculated by minimizing l(1) regularization-related objective function. Finally, a transform matrix that can preserve this relationship is obtained to project the mapped data into a low-dimensional space. The effectiveness of the proposed method is tested and compared with seven methods on four popular datasets.

• 出版日期2018-9
• 单位东北大学