graphlet kernels for large graph comparison

State-of-the-art graph kernels do not scale to large graphs with hundreds of nodes and thousands of edges. In this article we propose to compare graphs by counting graphlets, i.e., subgraphs with k nodes where k2 f3; 4; 5g. Exhaustive enumeration of all graphlets be- ing prohibitively expensive, we introduce two theoretically grounded speedup schemes, one based on sampling and the second one specif- ically designed for bounded degree graphs. In our experimental evaluation, our novel kernels allow us to eciently compare large graphs that cannot be tackled by existing graph kernels.