We consider the problem of sampling connected induced subgraphs of a given input graph G. Our first result is that an efficient algorithm to approximately sample connected induced subgraphs of a given size (the size is specified in the input) does not exist unless RP = NP. We then focus on the problem of approximately sampling connected induced subgraphs with a bias, more precisely we consider a distribution where the probability of a connected subgraph induced by S⊆ V(G) is proportional to λ|S|. When the input graph G has maximum degree d we identify a threshold λd=(d-1)(d-1)dd. For 0 < λ< λd there exists a trivial efficient sampler for the problem, and for λd< λ< 1 an efficient approximate sampler does not exist unless RP = NP. Finally, we show local Markov chains are unlikely to be effective at approximately sampling connected subgraphs.
CITATION STYLE
Read-McFarland, A., & Štefankovič, D. (2020). The Hardness of Sampling Connected Subgraphs. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 12118 LNCS, pp. 464–475). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-030-61792-9_37
Mendeley helps you to discover research relevant for your work.