Discriminant Uncorrelated Neighborhood Preserving Projections?

Dimensionality reduction is a crucial step for pattern recognition. Recently, a new kind of dimensionality reduction method, manifold learning, has attracted much attention.Among them, Neighborhood Preserving Projections (NPP) is one of the most promising techniques. In this paper, a novel manifold learning method called Discriminant Uncorrelated Neighborhood Preserving Projections (DUNPP), is proposed. Based on NPP, DUNPP takes into account the between-class information and designs a new difference-based optimization objective function with uncorrelated constraint. DUNPP not only preserves the within-class neighboring geometry, but also maximizes the between-class distance. Moreover, the features extracted via DUNPP are statistically uncorrelated with minimum redundancy, which is desirable for many pattern analysis applications. Thus it can obtain the stronger discriminant power. Experimental results on standard face database demonstrate the effectiveness of the proposed algorithm