On Different Stopping Criteria for Multi-objective Harmony Search Algorithms

Abstract

In evolutionary multi-objective optimization, an evolutionary algorithm is used to solve an optimization problem having multiple, and usually conflicting objective functions. Previous proposed approaches to solve multi-objective optimization problems include NSGA-II, MOEA/D, MOPSO, and MOHS/D algorithms. In our previous work, we enhanced the performance of MOHS/D using a hybrid framework with population diversity monitoring. The population diversity was measured every a predetermined number of iterations to either invoke local search or a diversity enhancement mechanism. In this work, two different stopping criteria are compared using four the HS hybrid frameworks we previously proposed. The stopping criteria compared are the moving average and MGBM. The experimental study is carried using the ZDT, DTLZ and CEC2009 benchmarks. The experimental results show that the moving average stopping criteria gives better results for the majority of the datasets.