Fine-grained parallel genetic algorithm: a stochastic optimisation method
Abubakr Muhammad, A Bargiela, G. R. Gnana King · 1997
This paper presents a fine-grained parallel genetic algorithm with mutation rate as a control parameter. The function of the mutation rate is similar to the function of temperature parameter in the simulated annealing [Lundy'86, Otten'89, and Romeo'85]. The parallel genetic algorithm presented here is based on a Markov chain [Kemeny'60] model. It has been proved that fine-grained parallel genetic algorithm is an ergodic Markov chain and it converges to the stationary distribution. 1. Introduction Parallel genetic algorithms are becoming more popular among researchers, due to the increased speed and efficiency [Muhlenbein'89]. There are two most popular parallel models of genetic algorithms. These are, distributed or coarse-grained model [Pettey'89 and Tanese'89], and massively parallel or fine-grained model [Manderick'89, Muhlenbein'91, and Tomassini'93]. The coarse-grained (distributed) parallel genetic algorithms assume the division of a large population into several subpopulations...