Supervised locality preserving projection for pattern classification
Abstract
Locality Preserving Projection (LPP) is a method for dimension reduction, which optimally preserves the neighborhood structure of the data set. This paper combines the label information with LPP, resulting Supervised Locality Preserving Projection (SLPP). SLPP projects the data into a lower dimensional subspace such that after the projection, the examples in different classes are located in different clusters, and the clusters are separated as far as possible. Thus, the projected samples by SLPP are better suited for classification than LPP. The experiments on face and handwritten digits classification verified that the same classifier can achieve a better performance with SLPP compared to LPP, which demonstrate that SLPP is more efficient in extracting discriminative information for pattern classification.
Department(s)
Mathematics
Document Type
Conference Proceeding
Publication Date
12-1-2010
Recommended Citation
Zheng, Songfend, Weixiang Liu. 2010. "Supervised locality preserving projection for pattern classification." In International Conference on Artificial Intelligence and Pattern Recognition, Orlando, FL, July 12-14, 2010, 19-23.
Journal Title
International Conference on Artificial Intelligence and Pattern Recognition 2010, AIPR 2010