基于圆形信赖域的改进和声搜索算法

Improved harmony search algorithm based on circular trust region

针对标准和声搜索(HS)算法易陷入局部最优、收敛精度不高的不足,提出了一种基于圆形信赖域(CTR)的新型和声搜索算法——CTRHS。该算法运用逐双音调一次性产生方式,在记忆思考环节交互式地采取面向圆形信赖域的集约化思考操作,在双音调微调环节利用当前和声记忆库中的最好或最差和声来确定微调带宽,并且以新生成和声直接替换当前和声记忆库中最差和声来实现和声记忆库的更新。通过在9种标准测试函数上对CTRHS算法进行实验验证和算法性能对比,结果表明CTRHS算法在解质量、收敛性能上优于文献中已报道的7种HS改进算法,且当和声记忆库规模(HMS)、和声记忆库思考率(HMCR)分别取5和0.99时,它能表现出更佳的全局优化性能。

Concerning the drawbacks of trapping in local optimal solutions and low convergence accuracy of standard Harmony Search（ HS） algorithm, a new harmony search algorithm based on Circular Trust Region（ CTR）, named as CTRHS,was proposed. CTRHS adopted the one-off generation mode of two pitches. Intensive considerations within the circular trust region were interactively conducted in its memory considering process. Adjustment bandwidth was determined by means of the best or worst harmony vector of current Harmony Memory（ HM） during the adjusting process of double pitches. The update of HM was achieved by replacing the worst harmony in current HM with the newly generated harmony. Computational experiments were conducted upon 9 benchmark functions to validate the performance of CTRHS. As demonstrated in the results, CTRHS outperforms other 7 reported HS variants in terms of solution quality and convergence efficiency. Moreover, when the parameters of Harmony Memory Size（ HMS） and Harmony Memory Considering Rate（ HMCR） are respectively equal to 5 and0. 99, it has better performance in searching the global optimal solutions.