Exploiting graph embedding in support vector machines

In this paper we introduce a novel classification framework that is based on the combination of the support vector machine classifier and the graph embedding framework. In particular we propose the substitution of the support vector machine kernel with sub-space or sub-manifold kernels, that are constructed based on the graph embedding framework. Our technique combines the very good generalization ability of the support vector machine classifier with the flexibility of the graph embedding framework resulting in improved classification performance. The attained experimental results on several benchmark and real-life data sets, further support our claim of improved classification performance.