A novel nonlinear dimensionality reduction approach for face recognition
Tao Wu · Journal of Circuits and Systems · 2009
Locally linear embedding(LLE) is one of the recently proposed manifold learning algorithms for nonlinear dimensionality reduction,which has demonstrated promising results in visualizing high dimensional data.However,the LLE lacks a parametric mapping between the observation and the low-dimensional output,In addition,since it is developed based on minimizing the reconstruction error,it may not be optimal from classification viewpoint.In this paper,we present a novel nonlinear dimensionality reduction approach for face recognition by fusion of LLE and LDA,Experiments on three public available face databases show the advantages of our proposed novel approach.