Synergies between evolutionary computation and probabilistic graphical models

Evolutionary computation and probabilistic graphical models have been two of the most lively research topics in artificial intelligence during the 1990s. However, with a few exceptions, each topic has been treated and developed separately. Recently, there has been an attempt to leverage the advances of one field into the other. This special issue is focused on cross-fertilization aspects between probabilistic graphical models and evolutionary computation. The six papers included in the number cover two types of synergies between both paradigms. On the one hand probabilistic graphical models are used to develop a new paradigm for evolutionary computation, named estimation of distribution algorithms (EDAs) [1]. This new class of algorithms generalizes genetic algorithms by replacing the crossover and mutation operators with the learning and sampling from the probability distribution of the best individuals of the population at each iteration of the algorithm. Working in such a way, the relationships between the variables involved in the problem domain are explicitly and effectively captured and exploited. On the other hand a recently proposed metaheuristic named ant colony optimization is used to learn Bayesian network structures from data. The paper by M€ uhlenbein and Mahnig presents a theory of population based optimization methods using approximations of search distributions. The authors prove the convergence of the search distributions to the global optima when the search distribution is a Boltzman distribution at each step and the size of the population is large enough. Also the relation between these optimization methods and those used in statistical physics is discussed. An application of the proposed method to the bipartitioning of large graphs is presented. Baluja s paper demonstrates how a priori knowledge of parameter dependencies, even incomplete knowledge, can be incorporated to efficiently obtain International Journal of Approximate Reasoning 31 (2002) 155–156 www.elsevier.com/locate/ijar