Manifold Regularized Proximal Support Vector Machine via Generalized Eigenvalue

Abstract

Proximal support vector machine via generalized eigenvalue (GEPSVM) is a recently proposed binary classification technique which aims to seek two nonparallel planes so that each one is closest to one of the two datasets while furthest away from the other. In this paper, we proposed a novel method called Manifold Regularized Proximal Support Vector Machine via Generalized Eigenvalue (MRGEPSVM), which incorporates local geometry information within each class into GEPSVM by regularization technique. Each plane is required to fit each dataset as close as possible and preserve the intrinsic geometric structure of each class via manifold regularization. MRGEPSVM is also extended to the nonlinear case by kernel trick. The effectiveness of the method is demonstrated by tests on some examples as well as on a number of public data sets. These examples show the advantages of the proposed approach in both computation speed and test set correctness.

Keywords

• Support vector machines
• Generalized eigenvalues
• Locality preserving projections
• Manifold regularization