一种从目标空间反向引导种群进化的进化算法

An evolutionary algorithm to reverse-guide population evolution from objective space

目前进化算法大多是通过解从决策空间到目标空间的映射,来判断解的质量。针对约束多目标优化问题,将极限学习机代理模型与不可行解存档方法相结合,提出一种通过目标向量反向预测来引导决策空间种群进化的算法。在CTP和TYPE系列的测试问题上进行了HV度量、IGD度量的性能测试。与几种经典的算法比较,该算法在大多情况下都表现出具有竞争力的性能,且在高难度问题下比其他算法表现更好。

Most of the current evolutionary algorithms judge the quality of the solution by mapping the solution from the decision space to the objective space.For the constrained multi-objective optimization problem,combining the extreme learning machine surrogate model with the infeasible solution archiving method,an algorithm to guide the evolution of the decision space according to backward prediction of the objective vector is proposed.The comparison of HV indicators and IGD indicators on CTP and TYPE test series shows that the proposed algorithm performs better than other classical algorithms in most cases,especially in high complexity problems.