改进的遗传算法选择算子

Improved selection operator of genetic algorithm

为了进一步提高遗传算法选择算子的选优能力,本文在传统轮盘赌的基础上提出了一种基于排序的多轮轮盘赌选择算子,加入了多轮转盘的思想,并利用对上代种群进行排序增加优秀个体的选择概率,从而在提高了算子的选优能力同时也减少了随机性所产生的误差;随后本文将此算子与最佳个体保存法的思想相结合,进一步提出了无放回的基于排序的多轮轮盘赌选择算子,达到了既能够选出最好个体又能够保证种群多样性的效果.实验表明,与传统轮盘赌算子相比较,新方法能够有效地提高遗传算法的收敛速度.

In order to improve optimal ability of selection operator of Genetic Algorithm, a new population- sorted multi- roulette- wheel selection operator （PSMRWS） is provided by studying traditional RWS. Based on the idea of multi-wheel, new algorithm reduces the selected error generated by the randomicity of roulette-wheel. Besides, better individual selection probability is promoted greatly by sorting population before selecting. Then based on result of PSMRWS, applying the idea of eliflist model, another algorithm population-sorted multi-roulette wheel selection with replacemen （PSMRWSR） is proposed, which can make better individuals being chosen to the next generation directly and in the meantime, diversity of population is kept. Experiments show that new selection operators can make GA be of prominently better convergent performance than the traditional RWS.