一种求解昂贵黑箱优化问题的多代理辅助进化算法

A Multiple Surrogates Assisted Evolutionary Algorithm for Expensive Black-Box Optimization

[目的]提出一种求解昂贵黑箱优化问题的多代理辅助进化算法。[方法]对进化采样辅助优化算法进行改进,将全局搜索中每代的进化操作进行10次,以降低求解的不稳定性;并对全局搜索与局部搜索的转换采用自适应距离准则判断,从而提高求解的精度。[结果]得到了新的昂贵黑箱优化问题的多代理辅助进化算法。[结论]使用22个测试问题对新算法的数值结果进行评估,结果表明新算法与进化采样辅助优化算法相比优势明显。

[Purposes]To propose a multiple surrogates assisted evolutionary algorithm for solving expensive black-box optimization problems.[Methods]The ESAO algorithm is improved by performing 10 generations of evolutionary operations in global search to reduce the instability of the solution,and using the adaptive distance criterion to judge the conversion between the global search and the local search,so as to improve the accuracy of the solution.[Findings]A new multiple surrogates assisted evolutionary algorithm for expensive black-box optimization problems is obtained.[Conclusions]The numerical results of the new algorithm are evaluated by using 22 test problems,and the results show that the new algorithm has significant advantages over the ESAO algorithm.