基于进化梯度搜索的多目标混合优化算法  

Multi-objective hybrid optimization algorithm based on evolutionary gradient search

作　　者：刘皓阳[1] 祁荣宾[1] 

机构地区：[1]华东理工大学化工过程先进控制和优化技术教育部重点实验室,上海200237

出　　处：《计算机与应用化学》2016年第11期1197-1202,共6页Computers and Applied Chemistry

基　　金：国家自然科学基金优秀青年基金项目(61422303);国家自然科学基金面上项目(21276078);上海市重点学科建设项目资助(B504);上海市自然科学基金(15ZR1408900)

摘　　要：基于梯度信息的线性搜索法具有快速的收敛性,但易陷入局部最优。当优化目标不可解析时,基于梯度信息的算法便不易应用。多目标进化算法以其优秀的全局特性广泛地应用于多目标优化问题,但其算法比较耗时,收敛速度慢。对此,本文提出一种基于进化梯度搜索的多目标混合算法。首先,结合单目标优化中的爬山算法与进化梯度搜索法,得到一种多目标局部搜索算法。其次,在算法前期采用适应度概率策略选择个体进行局部搜索。最后,在非支配集个体数达到种群个体数后,应用多目标进化算法保证其分布性。通过ZDT系列测试函数验证并与NSGA-II及EGS-NSGA-II混合算法比较,结果显示本算法具有更好的全局性及收敛快速性。Linear search algorithm based on gradient information has fast convergence, but it is easy to fall into local optimum. When the optimization goal is not differentiable, the algorithm based on gradient information will be no longer easy to apply. Multi-objective evolutionary algorithms are widely applied to multi-objective optimization problems for excellent global features, but this algorithm is time-consuming with slow convergence speed. For solving this problem, this paper proposes a hybrid multi-objective evolution algo- rithm based on gradient search. First of all, we promote a multi-objective local search algorithm compared with HCS and EGS algorithm which perform well in single objective problem. Secondly, at the early stage of algorithm, we selects individual species using fitness probability choosing strategy. Finally, when the non-dominated set number reaching the population individual number, multi-objective evolutionary algorithms return. According testing the algorithm through the ZDT series test function, and comparing with the NSGA-II and EGS-NSGA-II algorithms, results show that this algorithm has better global ability and fast convergence speed.

关 键 词：结构表征 荧光性能 发光材料 高分子 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]