M. Srinivas et Lm. Patnaik, ADAPTIVE PROBABILITIES OF CROSSOVER AND MUTATION IN GENETIC ALGORITHMS, IEEE transactions on systems, man, and cybernetics, 24(4), 1994, pp. 656-667
Citations number
25
Categorie Soggetti
Controlo Theory & Cybernetics","Computer Science Cybernetics","Engineering, Eletrical & Electronic
In this paper we describe an efficient approach for multimodal functio
n optimization using Genetic Algorithms (GAs). We recommend the use of
adaptive probabilities of crossover and mutation to realize the twin
goals of maintaining diversity in the population and sustaining the co
nvergence capacity of the GA. In the Adaptive Genetic Algorithm (AGA),
the probabilities of crossover and mutation, p(c) and p(m), are varie
d depending on the fitness values of the solutions. High-fitness solut
ions are 'protected', while solutions with subaverage fitnesses are to
tally disrupted. By using adaptively varying p(c) and p(m), we also pr
ovide a solution to the problem of deciding the optimal values of p(c)
and p(m), i.e., p(c) and p(m) need not be specified at all. The AGA i
s compared with previous approaches for adapting operator probabilitie
s in genetic algorithms. The sShema theorem is derived for the AGA, an
d the working of the AGA is analyzed. We compare the performance of th
e AGA with that of the Standard GA (SGA) in optimizing several nontriv
ial multimodal functions with varying degrees of complexity. For most
functions, the AGA converges to the global optimum in far fewer genera
tions than the SGA, and it gets stuck at a local optimum fewer times.
Our experiments demonstrate that the relative performance of the AGA a
s compared to that of the SGA improves as the epistacity and the multi
modal nature of the objective function increase. We believe that the A
GA is the first step in realizing a class of self organizing GAs capab
le of adapting themselves in locating the global optimum in a multimod
al landscape.