进化计算优化效率的定量评价  

Quantitative Evaluation of Optimization Efficiency for Evolutionary Computation

作　　者：温秀兰[1] 宋爱国[1] 

机构地区：[1]东南大学仪器科学与工程系

出　　处：《模式识别与人工智能》2005年第2期247-251,共5页Pattern Recognition and Artificial Intelligence

基　　金：国家自然科学基金(No.69875004);江苏省自然科学基金(No.BK2001402);东南大学-南瑞继保公司学位论文基金

摘　　要：进化计算实现时选用的进化算子不同、其计算的时间复杂度和寻优能力也不同。为了对进化计算优化效率进行定量评价,本文提出"优化平均截止时间"和"优化截止时间分布熵"两个概念,并以此作为评价准则。通过选用有代表性的基准测试函数,应用本文提出的评价准则对实数编码中典型的交叉算子进行大量测试,评价结果与理论分析完全一致,不仅证实本文提出的评价准则是正确有效的,而且为科学合理地选择算子及其算法提供理论依据。同时本文构造的算法可直接用于求解工程优化问题。Different evolutionary operators have different computation time complexity and different searching capability. In order to evaluate the optimization efficiency of evolutionary computation, two concepts, the optimization average truncated time and the distribution entropy of optimization truncated time, are proposed and unified as an evaluation criterion. By selecting several benchmark test functions and utilizing this criterion, several typical real-coded crossover operators are researched. The evaluation results are consistent with theoretical analysis. The conclusion not only shows that the evaluation criterion is correct and effective, but also provides a theoretical basis for selecting operators and algorithms scientifically and reasonably. In the meantime, the algorithm constructed in the paper is suitable for solving some engineering optimization problems.

关 键 词：进化计算 进化算子 优化效率 定量评价 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]