摘要
针对和声搜索算法在求解单峰值和多峰值函数寻优问题时存在收敛速度慢、精度低等问题,提出一种全局共享因子的和声搜索算法。在和声搜索算法基础上,引入全局共享因子思想,通过对和声搜索算法音调微调机制产生新和声方式,使改进算法能在初期利用较小的全局共享因子减弱音调微调带宽对最差和声的音调微调能力,迭代后期利用迅速增大到一定值后的全局共享因子来增强音调微调能力,最终实现全局收敛。从固定迭代次数和固定收敛精度两个角度分别对4个单、多峰值函数进行对比实验。结果表明:对多峰值Rastrigrin和Ackley函数,改进后的算法收敛精度、速度均优于和声搜索算法;对多峰值Griewank函数,改进后的算法在迭代次数10 000之后,其收敛精度、速度较和声算法无明显改进;改进算法对单峰值Rosenbrock函数的收敛精度、速度提升较为明显。
To solve the problem of slow convergence speed and low optimization precision of harmony search （HS） algorithm in solving problems of single peak function and multi-peak function, the har- mony search algorithm with global sharing factor （GSF-HS） was proposed. The global sharing factor i- dea was introduced based on the basic HS algorithm. The new tonality producing method was im- proved by GSF-HS algorithm in tonality adjustment mechanism of HS. In the initial stages, the smal-ler global sharing factor was used to reduce the tonality adjustment ability for the worst tonality by the improving algorithm, and in the late stages, the global sharing factor which was augmented to the fixed value was used to improve the tonality adjustment ability to achieve global convergence. There are two experiments including fixed global evolution times and fixed convergence precision by using four functions of single peak and multi-peak. For the multi-peak function Rastrigrin and Ackley, the results show the convergence precision and speed of GSF-HS algorithm is better than HS algorithm. For the multi-peak function Griewank, the convergence precision and speed of GSF-HS algorithm is maintained after 10000 iterative times compared with HS algorithm. Furthermore, for the single peak function Rosenbrock, the results show the convergence precision and speed of GSF-HS algorithm is obviously enhanced.
出处
《重庆理工大学学报（自然科学）》
CAS
2014年第2期82-86,共5页
Journal of Chongqing University of Technology：Natural Science
基金
甘肃农业大学盛彤笙科技创新基金资助项目(GSAU-STS-1322)
国家自然科学基金资助项目(61063028)
中国博士后科学基金资助项目(2013M542398)
兰州交通大学青年科学基金项目(2013032)
关键词
和声搜索算法
全局共享因子
音调微调
函数寻优
优化性能
harmony search algorithm
global sharing factor
tonality adjustment
function optimiza-tion
optimization performance
