Tree++: Truncated Tree Based Graph Kernels

Graph-structured data arise ubiquitously in many application domains. A fundamental problem is to quantify their similarities. Graph kernels are often used for this purpose, which decompose graphs into substructures and compare these substructures. However, most of the existing graph do not have property scale-adaptivity, i.e., they cannot at multiple levels granularities. Many real-world such as molecules exhibit structure varying To tackle problem, we propose a new kernel called Tree++ paper. At heart path-pattern kernel. The first builds truncated BFS tree rooted each vertex then uses paths from root every features represent graphs. can only capture similarity fine In order coarse granularities, incorporate concept super path it. contains trees vertices path. Our evaluation on variety demonstrates that achieves best classification accuracy compared with previous kernels.