An iterated local search algorithm for learning Bayesian networks with restarts based on conditional independence tests

Abstract

A common approach for learning Bayesian networks (BNs) from data is based on the use of a scoring metric to evaluate the fitness of any given candidate network to the data and a method to explore the search space, which usually is the set of directed acyclic graphs (DAGs). The most efficient search methods used in this context are greedy hill climbing, either deterministic or stochastic. One of these methods that has been applied with some success is hill climbing with random restart. In this article we study a new algorithm of this type to restart a local search when it is trapped at a local optimum. It uses problem-specific knowledge about BNs and the information provided by the database itself (by testing the conditional independencies, which are true in the current solution of the search process). We also study a new definition of neighborhood for the space of DAGs by using the classical operators of arc addition and arc deletion together with a new operator for arc reversal. The proposed methods are empirically tested using two different domains: ALARM and INSURANCE.

Iterated Local Search: Framework and Applications

“…ILS is used as a local search procedure within a GRASP approach by Ribeiro and Urrutia for tackling the mirrored traveling tournament problem [74]. Very high performing ILS algorithms have also been proposed for problems such as maximum clique [50], image registration [24], some loop layout problems [10], linear ordering [19,78], logistic network design problems [22], a capacitated hub location problem [75], Bayesian networks structure learning [25], and minimum sum-of-squares clustering [64].…”

Validation of a bayesian belief network representation for posterior probability calculations on national crime victimization survey

“…Results suggest that a number of BBN models performed exceptionally well as classifiers for the ''Victimization'' attribute. In fact, all listed versions of the local hill climbers and local K2 search algorithms (de Campos et al 2003;Madden 2003) led to classification performances with 97% or better accuracy. It is noted that the build time for these algorithms are on the same order and therefore, the build time was not considered as a further significant factor in performance comparisons.…”

Untitled

“…One of them is a greedy hill-climbing search with random restarts [27]. A random network model is selected for each restart.…”