基于PK模型的一种自适应遗传算法研究

Research on adaptive genetic algorithm based on PK model

遗传算法可以被理解为在逐代演化的过程中,适应性强的个体或种群具有更高的生存可能性的一种并行搜索算法。提出了基于PK竞争策略的遗传算法(Player Killing Genetical Algorithm,PKGA),其核心思想在于通过PK赛式的竞争筛选,直至剩下一个全程最优的个体即为全局最优解。通过对全程最优解的即时检测,同时配合交叉率与变异率在个体粒度上自适应地动态调整,算法能很好地避开局部极值点并减少进化过程中的退化现象。这种PK竞争筛选策略保证了算法较高的搜索效率和较强的鲁棒性。仿真实验证明,算法在应对早熟问题和退化现象及收敛效率等方面明显优于传统的标准遗传算法。

As parallel searching and optimization methods,Genetic algorithms promise that the individuals or populations with better adaptability have a higher possibility to survive in the process of evolution.According to which,an adaptive genetic algorithm based on PK model（Player Killing Genetieal Algorithm,PKGA） is proposed.Its core idea is that the best individual,as the global optimal solution,will survive by PK competition at the end of the evolution.With the real-time detection of the global optimal solution and the adaptive and dynamic adjustment of cross-rate and mutation-rate in individual size,the PKGA is able to overcome GA deception problem and reduce the degradation phenomenon of evolution.The PK competitive strategy ensures that PKGA is an efficient and robust searching and optimization method.Experiments show that PKGA is superior to traditional simple genetic algorithm.