Abstract:
Graphlet enumeration is known to be a fundamental yet challenging task in graph analysis. While significant progress has been made for undirected graphs, the enumeration of these substructures in directed graphs remains underexplored. There are critical inconsistencies in current graph processing systems and there is no single reference implementation for triangle counting on directed graphs, resulting in different systems reporting inconsistent triangle counts for the same dataset. In this work, we focus on efficient techniques for enumerating 3-cycles (triangles) and 4-cycles in large-scale graphs, with current emphasis on 3-cycle enumeration.