Learning spectral graph mapping for classification

Abstract

Nonlinear multi-classification has been a popular task in machine learning recently. In this paper, we propose a nonlinear multi-classification algorithm named Supervised Spectral Space Classifier (S3C), S3C integrates the discriminative information into the spectral graph mapping and transforms the input data into the low-dimensional supervised spectral space. S3C not only enables researchers to examine the mapped data in its supervised spectral space, but also can be directly applied to multi-classification problems. Experimental results on synthetic and real-world datasets demonstrate that S3C outperforms the state-of-the-art nonlinear classifiers SVM.