CAJ | 학술논문

传统演化算法在演化过程中种群会逐渐收敛，导致多样性丧失，一旦环境发生变化无法快速跟踪到变化的最优解。为了使演化算法能更好处理动态环境下优化问题，将预测和记忆结合，提出了一种处理有限离散环境状态变化的动态优化问题的遗传算法(MMGA)。该算法使用马氏链预测未来各个环境状态发生的概率，并结合存储在记忆中与环境状态有关的最优个体和等位基因分布向量产生适合未来环境的新种群。通过对动态优化测试问题进行实验，验证了该算法的离线性能优于其他三种经典算法。
전통연화산법재연화과정중충군회축점수렴，도치다양성상실，일단배경발생변화무법쾌속근종도변화적최우해。위료사연화산법능경호처리동태배경하우화문제，장예측화기억결합，제출료일충처리유한리산배경상태변화적동태우화문제적유전산법(MMGA)。해산법사용마씨련예측미래각개배경상태발생적개솔，병결합존저재기억중여배경상태유관적최우개체화등위기인분포향양산생괄합미래배경적신충군。통과대동태우화측시문제진행실험，험증료해산법적리선성능우우기타삼충경전산법。
The population of traditional evolutionary algorithm would converge gradually and the diversity of population would be lost in the course of evolution. Once the environment changes and could not track fast the change’s optimum solution. To allow the evolutionary algorithm to deal with the optimal issue better in the dynamic environment, this paper presents a genetic algorithm (MMGA) that deals with dynamic optimal issue with limited discrete environment state changing. This proposed algorithm uses Markov chain to predict the probabilities of various environmental state occurring, and generates new population that is fit for the future environment by combining the optimal entity that is stored in memory and related to the environment with the allele distribution vectors. By experimenting on the test of dynamic optimization, it proves that the off-line performance of this proposed algorithm is superior to other three classical algorithms.