Mine weighted network motifs via Bayes' theorem

Motif mining could recognize mecroscopic patterns of simple networks, especially with binary edges. Nevertheless, how to screen weighted motifs out of weighted networks is still a challenge. The difficulty mainly lies in two aspects, i.e. weight abstraction and null model enumeration. In this paper, we firstly model the heterogeneous weights as a fuzzy number problem, and normalize them into binary packages; then we neglect the time-consuming Markov process and discriminate weighted subgraphs by Bayes' equation. The simulation shows that our method is feasible and applicable.