Computing Posterior Probabilities of Structural Features in Bayesian Networks

We study the problem of learning Bayesian network structures from data.\nKoivisto and Sood (2004) and Koivisto (2006) presented algorithms that can\ncompute the exact marginal posterior probability of a subnetwork, e.g., a\nsingle edge, in O(n2n) time and the posterior probabilities for all n(n-1)\npotential edges in O(n2n) total time, assuming that the number of parents per\nnode or the indegree is bounded by a constant. One main drawback of their\nalgorithms is the requirement of a special structure prior that is non uniform\nand does not respect Markov equivalence. In this paper, we develop an algorithm\nthat can compute the exact posterior probability of a subnetwork in O(3n) time\nand the posterior probabilities for all n(n-1) potential edges in O(n3n) total\ntime. Our algorithm also assumes a bounded indegree but allows general\nstructure priors. We demonstrate the applicability of the algorithm on several\ndata sets with up to 20 variables.\n

Paper

Similar papers

© 2026 NYSGPT2525 LLC