BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20121114T183000Z
DTEND:20121114T184500Z
LOCATION:155-F
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: Bayesian networks (BNs) are probabilistic graphical models which have been used to model complex regulatory interactions in the cell (gene networks). However, BN structure learning is an NP-hard problem and both exact and heuristic methods are computationally intensive with limited ability to produce large networks. To address these issues, we developed a set of parallel algorithms. First, we present a communication efficient parallel algorithm for exact BN structure learning, which is work- and space-optimal, and exhibits near perfect scaling. We further investigate the case of bounded node in-degree, where a limit d on the number of parents per variable is imposed. We characterize the algorithm's run-time behavior as a function of d. Finally, we present a parallel heuristic approach for large-scale BN learning, which aims to combine the precision of exact learning. We evaluate the quality of the learned networks using synthetic and real gene expression data.
SUMMARY:Parallel Algorithms for Bayesian Networks Structure Learning with Applications to Gene Networks
PRIORITY:3
END:VEVENT
END:VCALENDAR