A hybrid methodology for learning belief networks: BENEDICT

Lecture Notes in Computer Science

Huete

2001

Self Cite

Abstract. Although we can build a belief network starting from any ordering of its variables, its structure depends heavily on the ordering being selected: the topology of the network, and therefore the number of conditional independence relationships that may be explicitly represented can vary greatly from one ordering to another. We develop an algorithm for learning belief networks composed of two main subprocesses: (a) an algorithm that estimates a causal ordering and (b) an algorithm for learning a belief network given the previous ordering, each one working over different search spaces, the ordering and dag space respectively.

“…It is part of a family of algorithms [2,3] that share a common methodology for learning belief networks, which we have called BENEDICT.…”

Section: Learning Belief Network With a Given Orderingmentioning

confidence: 99%

Section: Learning Belief Network With a Given Orderingmentioning

confidence: 99%

See 1 more Smart Citation

The Search of Causal Orderings: A Short Cut for Learning Belief Networks

Acid

Lecture Notes in Computer Science

Huete

2001

Self Cite

“…Algorithms for learning (the structure of) BNs have been studied, basically from two points of view: Methods based on conditional independence tests [5,6,7,22,23] and methods based on a scoring metric optimization [12,16,17]. This classification is not exhaustive and/or strict, there also exist algorithms that use a combination of these two methods [1,2,13,21]. In this paper we only consider learning methods based on a scoring metric.…”

Section: Introductionmentioning

confidence: 99%

“…Moreover, it is also known that, by providing a good ordering of the variables, the learning algorithms become more efficient and accurate. In fact, there is a number of algorithms that need to use such an ordering [1,2,7,10,12]. Therefore, our proposal is to develop learning methods that carry out the search process in the space of the orderings instead of the space of dags.…”

Section: Introductionmentioning

confidence: 99%

Stochastic Local Algorithms for Learning Belief Networks: Searching in the Space of the Orderings

Lecture Notes in Computer Science

Puerta

2001

Self Cite

Abstract. An important type of methods for learning belief networks from data are those based on the use of a scoring metric, to evaluate the fitness of any given candidate network to the data base, and a search procedure to explore the set of candidate networks. In this paper we propose a new method that carries out the search not in the space of directed acyclic graphs but in the space of the orderings of the variables that compose the graphs. Moreover, we use a new stochastic search method to be applied to this problem, Variable Neighborhood Search. We also experimentally compare our methods with some other search procedures commonly used in the literature.

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

Fernández‐Luna

Puerta

2003

Int. J. Intell. Syst.

Self Cite

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.