Parallel BMDA with an aggregation of probability models

Abstract

The paper is focused on the problem of aggregation of probability distribution applicable for parallel Bivariate Marginal Distribution Algorithm (pBMDA). A new approach based on quantitative combination of probabilistic models is presented. Using this concept, the traditional migration of individuals is replaced with a newly proposed technique of probability parameter migration. In the proposed strategy, the adaptive learning of the resident probability model is used. The short theoretical study is completed by an experimental works for the implemented parallel BMDA algorithm (pBMDA). The performance of pBMDA algorithm is evaluated for various problem size (scalability) and interconnection topology. In addition, the comparison with the previously published aBMDA [24] is presented.

Distributed Estimation of Distribution Algorithms for continuous optimization: How does the exchanged information influence their behavior?

“…A different strategy is to use a specific operator that takes into account the structure of the models to combine them appropriately. This method has the disadvantage that, with complex models, the combination is not trivial and, in some cases, can be very inefficient [20]. In the simple UMDA g models, a straightforward and efficient approach that can be carried out, named UMDA g combination method, is to combine the means and variances vector, i.e.…”

“…In order to compute the value of fi, the strategy that has been followed in previous studies [10,20] is the so-called adaptative learning strategy. This method computes the fi values based on the quality of the population associated to each model.…”

“…In order to analyze the effects of the migration strategies, several island configurations instances were compared against each other. The selection of the parameter values, such as the learning model, was based on the values used in previous studies with dEDAs [12,20] and was extended with additional values to obtain a wider view. Table 2 shows the different parameter values used throughout the experiments.…”

“…Migration of individuals is a classic approach and has proven to obtain successful results in these and other Evolutionary Algorithms [2,4,9,28]. In addition, migration of models was explicitly developed for the distributed Estimation of Distribution Algorithms [1,11,12,19,20].…”

Variable Classification-based Co-estimation of Distribution Algorithm Assisted by a Cooperative Particle Swarm Optimizer

“…migration operation) among different sub-groups at a certain evolutionary generation to generate new population and evolve the next generation on the basis of this cycle until the algorithm termination condition is satisfied. A migration operation is proven to be effective for EDAs and is known to yield good results [5,6,7,8,9].…”