Searching a multivariate partition space using MAX-SAT

Authors:
Silvia Liverani;James Cussens;Jim Q. Smith
Affiliations:
Department of Mathematics, University of Bristol, Bristol, UK;York Centre for Complex Systems Analysis & Department of Computer Science, University of York, York, UK;Department of Statistics, University of Warwick, Coventry, UK
Venue:
CIBB'09 Proceedings of the 6th international conference on Computational intelligence methods for bioinformatics and biostatistics
Year:
2009

Citing 2
Cited 0

Bayesian Treed Models

Machine Learning
UBCSAT: an implementation and experimentation environment for SLS algorithms for SAT and MAX-SAT

SAT'04 Proceedings of the 7th international conference on Theory and Applications of Satisfiability Testing

Quantified Score

Hi-index	0.00

Visualization

Abstract

Because of the huge number of partitions of even a moderately sized dataset, even when Bayes factors have a closed form, a comprehensive search for the highest scoring (MAP) partition is usually impossible. Therefore, both deterministic or random search algorithms traverse only a small proportion of partitions of a large dataset. The main contribution of this paper is to encode the formal Bayes factor search on partitions as a weighted MAX-SAT problem and use well-known solvers for that problem to search for partitions. We demonstrate how, with the appropriate priors over the partition space, this method can be used to fully search the space of partitions in smaller problems and how it can be used to enhance the performance of more familiar algorithms in large problems. We illustrate our method on clustering of time-course microarray experiments.