Geometry preserving graph embedding based on probabilistic collaborative representation
(1. Artificial Intelligence Key Laboratory of Sichuan Province, Sichuan University of Science and Engineering, Zigong, Sichuan 643002, China； 2. School of Computer Science and Communication Engineering, Jiangsu University, Zhenjiang, Jiangsu 212013, China)
Abstract：To fully capture the geometric information of data and enhance the power of pattern discrimination, a dimension reduction method of probabilistic collaborative representation geometry preserving graph embedding (PCRGPGE) was proposed based on the ability of probabilistic collaborative representation for capturing the discriminant and geometric information of data. In the new method, the similar highdimensional samples had the similar probabilistic collaborative representations, which were also similar in the subspace through graph embedding, so that the lowdimensional latent structure of highdimensional data was well preserved in the subspace. In the PCRGPGE, the high dimensional data was given by the probabilistic collaborative representation and reconstructed to obtain the natural discriminant and geometric information. The intraclass and interclass graph constructions were used to discover the discriminant information and the geometric distributions of data. The experiments were conducted to compare PCRGPGE with ten stateoftheart graph embedding methods such as PCA, LDA and LPP on two public PIE29 and IMM face image classification data sets. The results show that the proposed PCRGPGE is a promising dimension reduction method, which can well preserve the intrinsic structure of highdimensional data and enhance the pattern discrimination in the lowdimensional space.
