| 改进进化策略及其在神经网络训练中的应用 |
| |
| 引用本文： | 柯晶,姜静,李歧强.改进进化策略及其在神经网络训练中的应用[J].计算机工程与应用,2006,42(4):68-70,141. |
| |
| 作者姓名： | 柯晶 姜静 李歧强 |
| |
| 作者单位： | 山东大学控制科学与工程学院,济南,250061;海军航空工程学院自动控制系,烟台,264001 |
| |
| 基金项目： | 教育部科学技术研究项目 |
| |
| 摘 要： | 进化策略是一类策略参数自适应进化算法。文章提出了一种改进进化策略(MES),MES采用基于个体排序的随机自适应Gaussian-Cauchy混合变异策略,将Gaussian和Cauchy变异算子结合起来以达到全局探索和局部搜索之间的动态平衡。此外,MES还使用重组算子以进一步提高算法的性能。将该算法用于多层前向神经网络训练,数值仿真结果显示了该算法的有效性。
|
| 关 键 词： | 进化算法 进化策略 神经网络 |
| 文章编号： | 1002-8331-（2006）04-0068-03 |
| 收稿时间： | 2005-06 |
| 修稿时间： | 2005-06 |
Modified Evolution Strategy and its Application to Neural Network Training |
| |
| Ke Jing,Jiang Jing,Li Qiqiang.Modified Evolution Strategy and its Application to Neural Network Training[J].Computer Engineering and Applications,2006,42(4):68-70,141. |
| |
| Authors: | Ke Jing Jiang Jing Li Qiqiang |
| |
| Affiliation: | 1.School of Control Science and Engineering,Shandong University,Jinan 250061; 2.Department of Automatic Control,Naval Aeronautical Engineering Institute,Yantai 264001 |
| |
| Abstract: | Evolution strategies are a class of evolutionary algorithms with self-adaptation of strategy parameters.A Modified Evolution Strategy(MES) is proposed.MES employs random adaptive Gaussian-Cauchy hybrid mutation strategy based on the rank order of individuals.Gaussian and Cauchy mutation operators are combined to achieve dynamic balance of global exploration and local search.Recombination operator is also adopted in MES to further enhance the performance.The proposed algorithm is applied to multilayer feedforward neural network training.Numerical simulation results show the effectiveness of the proposed algorithm. |
| |
| Keywords: | evolutionary algorithm evolution Strategies Neural Network |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
