不规则的Pareto前沿面问题,具有不连续、退化的、倒置等不规则形式,利用传统的优化算法往往不能达到最优值.针对此问题,通过改进RVEA算法,使用参考向量再生的方式丰富解集以充分遍历不规则Pareto前沿面,提出Dynamic-RVEA算法.该方法利用随机函数随机在目标向量的取值范围内生成新的参考向量来替换原来的单位参考向量,然后通过在DTLZ测试问题集上超体积指标(HV)实验结果性能对比.实验结果表明Dynamic-RVEA在超体积指标HV性能指标上表现突出.
The irregular Pareto front problems have irregular forms such as discontinuity,degradation,in-version,etc.The optimal value can not be achieved by traditional optimization algorithms.To solve this problem,the dynamic-RVEA algorithm was proposed by improving the RVEA algorithm and enriching the solution set by means of reference vector regeneration to fully traverse the irregular Pareto front surface.The method uses random function to generate new reference vector randomly within the value range of the target vector to replace the origi-nal unit reference vector,and then compares the performance of the HV experimental results on the DTLZ test problem set.The experimental results show that dynamic-RVEA has outstanding performance in HV performance index.