Graph Entropy from Closed Walk and Cycle Functionals

This paper presents an informational functional that can be used to characterise the entropy of a graph or network structure, using closed random walks and cycles. The work commences from Dehmer’s information functional, that characterises networks at the vertex level, and extends this to structures which capture the correlation of vertices, using walk and cycle structures. The resulting entropies are applied to synthetic networks and to network time series. Here they prove effective in discriminating between different types of network structure, and detecting changes in the structure of networks with time.