Parallel heterogeneous genetic algorithms for continuous optimization

Abstract

In this paper we address an extension of a very efficient genetic algorithm (GA) known as Hy3, a physical parallelization of the gradual distributed real-coded GA (GD-RCGA). This search model relies on a set of eight subpopulations residing in a cube topology having two faces for promoting exploration and exploitation. The resulting technique has been shown to yield very accurate results in continuous optimization by using crossover operators tuned to explore and exploit the solutions inside each subpopulation. We introduce here a further extension of Hy3, called Hy4, that uses 16 islands arranged in a hypercube of four dimensions. Thus, two new faces with different exploration/exploitation search capabilities are added to the search performed by Hy3. We analyze the importance of running a synchronous versus an asynchronous version of the models considered. The results indicate that the proposed Hy4 model overcomes the Hy3 performance because of its improved balance between exploration and exploitation that enhances the search. Finally, we also show that the async Hy4 model scales better than the sync one.

Customized crossover in evolutionary sets of safe ship trajectories

“…The choice of crossover operators has an impact on faster or slower convergence (Kowalczuk and Białaszewski, 2006) and therefore sometimes, e.g., in parallel genetic algorithms for continuous optimisation (Alba et al, 2004), special crossover operators promoting exploration or exploitation are introduced.…”

Application of parallel computing to stochastic parameter estimation in environmental models

“…optimization problems involving complex spatially distributed models, such as are frequently used in the field of earth science, this sequential implementation needs to be revisited (Abramson, 1991;Goldberg et al, 1995;Alba and Troya, 1999;Herrera et al, 1998;Vrugt et al, 2001;Alba et al, 2004;Eklund, 2004;de Toro Negro et al, 2004;Vrugt et al, 2004; amongst various others). Most computational time required for calibrating parameters in complex environmental models is spent running the model code and generating the desired output.…”

“…Parallel computing offers the possibility of solving computationally challenging optimization problems in less time than is possible using ordinary serial computing (Abramson, 1991;Goldberg et al, 1995;Alba and Troya, 1999;Herrera et al, 1998;Alba et al, 2004;Eklund, 2004;de Toro Negro et al, 2004;amongst various others). Despite these prospects, parallel computing has not entered into widespread use in the field due to difficulties with implementation and barriers posed by technical jargon.…”

A doubly distributed genetic algorithm for network coding

“…or other connected systems where local decision variables are to be specified for the optimal performance of the whole system. 2) Furthermore, the proposed framework of temporal axis distribution can be combined with, not only the pipelining methods considered in this paper, a fairly general class of state-of-the-art strategies for parallel management of populations and communication between populations (e.g., [2,14]), because it imposes essentially no constraint on the implementation of any such strategies except that there is slight temporal inconsistency between populations, which as shown in this paper may also have little effect to other strategies.…”