Supervised Nonlinear Dimensionality Reduction Based on Evolution Strategy
Mudasser Naseer, Shiyin Qin · 2013
Most of the classifiers suffer from the curse of dimensionality during classification of high dimensional image and non-image data. In this paper, we introduce a new supervised nonlinear dimensionality reduction (S-NLDR) algorithm called supervised dimensionality reduction based on evolution strategy (SDRES) for both image and nonimage data. The SDRES method uses the power of evolution strategy (ES) algorithm to find low dimensional embedding of high dimensional labeled data. The new algorithm makes the interclass dissimilarity larger than the intraclass dissimilarity while finding low dimensional embedding values. Simulation studies on some well-known benchmark datasets demonstrate that SDRES generally gives better results in dimensionality reduction and classification as compared to other famous S-NDLR methods such as WeightedIso, supervised S-Isomap, supervised locally linear embedding (SLLE), enhanced supervised locally linear embedding (ESLLE) and supervised local tangent space alignment (SLTSA).