摘要
为提高多目标粒子群算法(MOPSO)的收敛性与解集多样性,提出一种基于侧步爬山策略的混合多目标粒子群算法(H-MOPSO).通过建立局部搜索与粒子群优化的混合模型,在该模型中后期引入基于侧步爬山策略的局部搜索,周期性代替粒子群搜索并优化混合参数,使粒子根据距离前沿的远近朝下降或非支配方向搜索,加快粒子群收敛并改善其分布.同时采用非均匀变异算子和线性递减的惯性权重策略,避免算法早熟.通过标准测试函数的对比实验表明,该算法整体上比MOPSO、NSGA-II和MOEA/D具有更好的多样性与收敛性.
To improve the convergence and diversity performance of multi-objective particle swarm optimization { MOPSO }, a hybrid MOPSO ( H-MOPSO } based on hill climbing strategy with sidesteps is proposed. The hybrid model of local search and particle swarm optimization is firstly established. According to the model, the local search algorithm based on hill climbing strategy with sidesteps is introduced periodically in late period with optimized parameters, making particles search along descent direction while they are away from Pareto front, and search along the front while they are near Pareto front. The non-uniform mutation operator is used and the iner- tia weight is set to be linearly decreased to avoid premature. Simulation results indicate that the proposed algorithm has favorable per- formance comparing to MOPSO, NSGA-II and MOEA/D.
出处
《小型微型计算机系统》
CSCD
北大核心
2012年第12期2696-2702,共7页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61070135)资助
国家社会科学基金项目(10GBL095)资助
关键词
多目标优化
粒子群算法
侧步爬山策略
混合算法
multi-objective optimization
particle swarm optimization
hill climbing strategy with sidesteps
hybrid algorithm
作者简介
E-mail:qfy@zjut.edu.cn;wlp@zjut.edu.cn作者简介:王丽萍。女,1964年生,教授,博士生导师,研究方向为决策优化、智能系统等;
吴秋花,女,1987年生,硕士研究生,研究方向为商务智能;
邱飞岳,男,1965年生,教授,研究方向为科学与媒体技术、智能系统等;
吴裕市,男,1987年生,硕士研究生,研究方向为人工智能;
林思颖,女,1990年生,硕士研究生,研究方向为决策优化.