Stable and Discriminative Topological Graph Analysis
Abstract
We propose a novel method for topological analysis of unweighted graphs which is based on persistent homology. The proposed method maps the input graph to a complete weighted graph where the weighting function maps each edge to a value indicating the degree to which it belongs to a clique. The persistent homology of this weighted graph is subsequently computed to give a topological representation describing the topological features of the input graph plus their significance. A formal and experimental analysis of the proposed and existing methods for topological graph analysis is presented. Through this analysis, we find that the proposed method possesses the properties of being stable and performing accurate discrimination. Therefore this method can make accurate inferences regarding the topological features of a given graph. On the other hand, we find that the existing methods considered do not possess these properties making it difficult from them to make such inferences. These findings are experimentally demonstrated using a number of random and real world graphs.