Stopping Rules for Randomized Greedy Triangulation Schemes

Authors

  • Andrew Gelfand University of California, Irvine
  • Kalev Kask University of California, Irvine
  • Rina Dechter University of California, Irvine

DOI:

https://doi.org/10.1609/aaai.v25i1.8021

Abstract

Many algorithms for performing inference in graphical models have complexity that is exponential in the treewidth — a parameter of the underlying graph structure. Computing the (minimal) treewidth is NPcomplete, so stochastic algorithms are sometimes used to find low width tree decompositions. A common approach for finding good decompositions is iteratively executing a greedy triangulation algorithm (e.g. minfill) with randomized tie-breaking. However, utilizing a stochastic algorithm as part of the inference task introduces a new problem — namely, deciding how long the stochastic algorithm should be allowed to execute before performing inference on the best tree decomposition found so far. We refer to this dilemma as the Stopping Problem and formalize it in terms of the total time needed to answer a probabilistic query. We propose a rule for discontinuing the search for improved decompositions and demonstrate the benefit (in terms of time saved) of applying this rule to Bayes and Markov network instances.

Downloads

Published

2011-08-04

How to Cite

Gelfand, A., Kask, K., & Dechter, R. (2011). Stopping Rules for Randomized Greedy Triangulation Schemes. Proceedings of the AAAI Conference on Artificial Intelligence, 25(1), 1043-1048. https://doi.org/10.1609/aaai.v25i1.8021