Graph Clustering Using the Jensen-Shannon Kernel
This paper investigates whether the Jensen-Shannon divergence can be used as a means of establishing a graph kernel for graph classification. The Jensen-Shannon kernel is nonextensive information theoretic kernel which is derived from mutual information theory, and is defined on probability distributions. We use the von-Neumann entropy to calculate the elements of the Jensen-Shannon graph kernel and use the kernel matrix for graph classification. We use kernel principle components analysis (kPCA) to embed graphs into a feature space. Experimental results reveal the method gives good classification results on graphs extracted from an object recognition database.