Finding the optimal Bayesian network given a constraint graph

• Published in: PeerJ. Computer science Vol. 3; p. e122
• Main Authors: Schreiber, Jacob M., Noble, William S.
• Format: Journal Article
• Language: English
• Published: San Diego PeerJ. Ltd 03.07.2017; PeerJ, Inc; PeerJ Inc
• Subjects: Algorithms; Applied research; Artificial intelligence; Bayesian analysis; Bayesian network; Belief networks; Big data; Coding; Computation; Computer memory; Data analysis; Discrete optimization; Gene expression; Graphic methods; Graphs; International conferences; Knowledge; Learning; Machine learning; Methods; Optimization theory; Parallel processing; Parents & parenting; Random variables; Structure learning
• ISSN: 2376-5992; 2376-5992
• DOI: 10.7717/peerj-cs.122

Despite recent algorithmic improvements, learning the optimal structure of a Bayesian network from data is typically infeasible past a few dozen variables. Fortunately, domain knowledge can frequently be exploited to achieve dramatic computational savings, and in many cases domain knowledge can even make structure learning tractable. Several methods have previously been described for representing this type of structural prior knowledge, including global orderings, super-structures, and constraint rules. While super-structures and constraint rules are flexible in terms of what prior knowledge they can encode, they achieve savings in memory and computational time simply by avoiding considering invalid graphs. We introduce the concept of a “constraint graph” as an intuitive method for incorporating rich prior knowledge into the structure learning task. We describe how this graph can be used to reduce the memory cost and computational time required to find the optimal graph subject to the encoded constraints, beyond merely eliminating invalid graphs. In particular, we show that a constraint graph can break the structure learning task into independent subproblems even in the presence of cyclic prior knowledge. These subproblems are well suited to being solved in parallel on a single machine or distributed across many machines without excessive communication cost.