基于自适应学习的多目标粒子群优化算法

将进化算法应用于某些多目标优化问题时,采用增加种群规模和进化代数的方法往往耗费大量的目标函数计算开销,且达不到提高种群进化效率的目的,为此提出了一种基于自适应学习最优搜索方向的多目标粒子群优化算法。采用自适应惯性权值平衡算法的全局和局部搜索能力,采用聚类排挤方法保持Pareto非支配解集的分布均匀性,使用最近邻学习方法为每个粒子在Pareto非支配解集中寻找一个最优飞行目标来提高其收敛速度并保持粒子群搜索方向的多样性。实验结果表明,提出的算法可在显著地降低函数评估成本的前提下实现快速的搜索,并使粒子群均匀地逼近Pareto最优面。

Multi-objective particle swarm optimizationalgorithm based on self-adaptive learning

When evolutionary algorithm is applied to multi-objective optimization problems, it often requires a large population size and a large number of evolution generation. However, it consumed plenty of computation overhead of evaluating objective functions but just resulted in poor improvement of the search efficiency. This paper proposed a multi-objective particle swarm optimization algorithm based on self-adaptive learning of optimal search directions. The algorithm used the self-adaptive inertia weights to get the trade-off between the global and local search. And it used the clustering crowding to maintain the uniform distribution of the non-dominated Pareto solutions. And the algorithm incorporated the nearest neighbor rule to seek the best target in the non-dominated Pareto solutions to get the optimal flying direction for each particle, it sped the flying of single particle and kept the diversity of the flying directions for the particle swarm. The experimental results show that the algorithm can drive the particle swarm to approximate quickly and uniformly to the Pareto front, and decrease the evaluation cost of objective functions significantly.