摘要
在分析导致进化规划算法早熟原因的基础上,提出了一种基于遗忘策略的双群进化规划算法。在该算法中,进化在两个不同的子群间并行进行,其中一个子群使用遗忘策略不断淘汰和更新个体以实现在变量空间中足够分散的探索,另一个子群使用指数递减的高斯变异算子以实现在子群所在的局部尽可能细致搜索。通过种群重组实现子群间的个体与信息交流。基于典型算例的数字仿真证明该算法具有更好的全局收敛性,更快的收敛速度和更强的鲁棒性。
Based on the analysis of the premature convergence of traditional evolutionary programming, a forgetting strategy based bi-subgroup evolutionary programming （FSBEP） algorithm is proposed. In this algorithm, the evolution of two subgroups is parallelly performed by different mutation strategies. One subgroup eliminates and updates individuals to explore the variable separately enough by the forgetting strategy, and another one searches the local part using the exponential degressive Gaussian mutation operator. Information, together with individual, is exchanged when the population is reorganized. Simulations based on benchmarks confirm that the FSBEP algorithm is better than classical evolutionary programming algorithm in the aspects of global optimization, the convergence speed and the robustness.
出处
《数据采集与处理》
CSCD
北大核心
2005年第3期263-267,共5页
Journal of Data Acquisition and Processing
关键词
进化规划
双群
搜索
收敛
evolutionary programming
bi-subgroup
search
convergence
