多子群协同链式智能体遗传算法分析

针对遗传算法(genetic algorithm,GA)易出现搜索效率不高和早收敛现象,提出了一种多子群协同链式智能体遗传算法(multi-population agent genetic algorithm,MPAGA)。该算法采用多子群并行搜索模式、链式智能体结构,引入动态邻域竞争和正交交叉等策略,有效提高了算法性能。采用3个复杂多峰测试函数对算法进行优化性能测试结果表明,MPAGA比普通智能体遗传算法有较快的收敛速度,能有效防止早收敛现象。

Analysis of multi-population co-genetic algorithm for global numerical optimization

In order to improve the low optimization efficiency and the premature convergence of genetic algorithms(GA),a multi-population agent co-genetic algorithm with a chain-like agent structure(MPAGA) was developed.This algorithm adopted a multi-population parallel searching mode,a chain-like agent structure,dynamic neighborhood competition,and an orthogonal crossover strategy.Three functions were used to test this algorithm.The experimental results show that MPAGA obtains higher optimization precision and converges to the domain close to global optima with higher speeds than other improved GAs.