网络出版地址：http://www.cnki.net/kcms/detail/23.1538.TP.20150326.1016.004.html

动态调整策略改进的和声搜索算法

拓守恒,雍龙泉,邓方安

(陕西理工学院 数计学院，陕西 汉中 723001)

摘要：为了得到高维复杂问题的全局高精度最优解，提出一种动态调整策略，并用该策略改进和声搜索算法。算法选取和声记忆库中最差和声向量作为优化调整目标，随着迭代的进行，逐步降低决策变量的调整概率，该方法能够使得算法在全局探索能力和局部高精度开发能力之间实现平衡，有效提高了新和声更新最差和声的成功率。通过6个高维Benchmark测试函数的仿真结果表明，提出的动态调整策略能够有效提高和声搜索算法求解高维复杂优化问题的能力。

关键词：自适应调整策略；高维优化问题；和声搜索算法

DOI:10.3969/j.issn.1673-4785.201402019

中图分类号：TP391文献标志码：A

收稿日期：2014-02-21.网络出版日期：2015-03-26.

基金项目：国家自然科学基金资助项目(11401357)，陕西省教育厅科研资助项目(14JK1141)；汉中市科技局科研资助项目(2013hzzx-39);陕西理工学院科研资助项目(SLGKY 13-27).

作者简介：

中文引用格式：拓守恒,雍龙泉,邓方安.动态调整策略改进的和声搜索算法[J]. 智能系统学报， 2015, 10(2): 307-315.

英文引用格式：TUO Shouheng, YONG Longquan, DENG Fang’an. Dynamic adjustment strategy for improving the harmony search algorithm[J]. CAAI Transactions on Intelligent Systems, 2015, 10(2): 307-315.

Dynamic adjustment strategy for improving the harmony search algorithm

TUO Shouheng, YONG Longquan, DENG Fang’an

(School of Mathematics and Computer Science, Shaanxi University of Technology, Hanzhong 723001,China)

Abstract：A dynamic adjustment strategy is used to improve the harmony search algorithm for solving high-dimensional multimodal global optimization problems. It chooses the worst harmony vector from the harmony memory (HM) as an optimization objective vector. With the process of iteration, the adjustment probability of decision variables is reduced step by step. It can achieve the balance effectively between the global exploration powers and local exploitation competence， and can increase the success rate of evolution. Finally, the experimental results of 16 high-dimension benchmark functions demonstrated that the proposed method can enhance the performance and robustness of the harmony search algorithm obviously in solving large scale multimodal optimization problems.

Keywords：adaptive adjustment strategy; high-dimensional optimization problems; harmony search algorithm

通信作者：拓守恒.tuo_sh@126.com.

近年来，随着社会的发展和大数据时代的到来，人们对科技产品的能耗和性能要求越来越高，使得产品的设计遇到了极大的挑战。许多产品的设计需要考虑很多的设计要求，并要使其产品整体设计能够达到最优,这些问题都可转化为大规模复杂优化问题(optimization problems,OP)。对于OP问题，近年来，研究者将关注的焦点集中在模拟自然的启发式(meta-heuristics)优化方法[1-9]等。……