摘要
The personal best is an interesting topic, but little work has focused on whether it is still efficient for multiobjective particle swarm optimization. In dealing with single objective optimization problems, a single global best exists, so the personal best provides optimal diversity to prevent premature convergence. But in multiobjective optimization problems, the diversity provided by the personal best is less optimal, whereas the global archive contains a series of global bests, thus provides optimal diversity. If the algorithm excluding the personal best provides sufficient randomness, the personal best becomes worthless. Therefore we propose no personal best strategy that no longer uses the personal best when the global archive exceeds the population size. Experimental results validate the efficiency of our strategy.
The personal best is an interesting topic, but little work has focused on whether it is still efficient for multiobjective particle swarm optimization. In dealing with single objective optimization problems, a single global best exists, so the personal best provides optimal diversity to prevent premature convergence. But in multi- objective optimization problems, the diversity provided by the personal best is less optimal, whereas the global archive contains a series of global bests, thus provides optimal diversity. If the algorithm excluding the personal best provides sufficient randomness, the personal best becomes worthless. Therefore we propose no personal best strategy that no longer uses the personal best when the global archive exceeds the population size. Experimental results validate the efficiency of our strategy.
基金
the Research Funds of ShanghaiMunicipal Science and Technology Commission(No.12511502902)
the National Natural ScienceFoundation of China(No.61375053)
