This software package provides a framework for calculating similarity between neighborhoods rooted at two vertices of interest in a labeled graph (undirected or directed). The list of available similarity functions includes: cumulative random walk, standard random walk, standard graphlet kernel, edit distance graphlet kernel, label substitution graphlet kernel and edge indel graphlet kernel. The graphlet kernel framework can be used for vertex (node) classification in graphs, kernel-based clustering, or community detection.
If you use this framework, please cite the following paper:
Lugo-Martinez J. and Radivojac P. "Generalized Graphlet Kernels for Probabilistic Inference in Sparse Graphs", Network Science, 2014.
Be the first to post a review of Graphlet kernel framework!