摘要
A supervised multi-manifold based discriminant analysis algorithm is presented. Sparse local scatter is derived for within-class compactness description while multi-class nonparametric scatter is used to characterise between-class separability. By preserving the within-class compactness and maximising the between-class separability simultaneously, the algorithm seeks for the optimal projection matrix to identify the underlying manifold structure of a multi-class dataset. Experimental results on data visualisation and face recognition demonstrate that the presented method is robust and efficient.
- 出版日期2013-1-31
- 单位中国人民解放军国防科学技术大学