摘要
对于高维复杂函数优化问题,经典的优化算法存在着初始点敏感、局部收敛等问题;而模拟退火算法等智能算法则有着计算成本高昂、算法早熟等缺陷。NFL定理犤1犦预示了混合优化策略是解决实际优化问题的最好途径。该文融合了模拟退火算法和经典算法的优点,设计了高维复杂函数混合模拟退火优化策略。混合优化策略具有模拟退火算法的全局收敛性,同时引入强局部收敛经典算法作为模拟退火算法的精英个体提高算子,提高了模拟退火算法局部开采能力,加快了收敛速度。数值仿真计算结果表明,混合模拟退火策略求解高维复杂函数的性能大大优于单一算法,具有强鲁棒性、高收敛速度和高精度等优点。该文的算法设计思想对于解决实际问题有较好的借鉴意义。
For optimizing high-dimension complex functions,the classical optimization algorithms have the problems of high sensitivity to initial guess and local convergence,while the intelligent optimization algorithms such as Simulated Annealing(SA)have the limitations of high computational costs and premature.NFL theorem has indicated that hybrid optimization approach is the best method for solving the practical problems.The hybrid approach for optimizing high-dimension complex functions is proposed in this paper,which combines the merits of SA with that of the classical algorithms.The hybrid approach has global convergence property of SA,and adopts high -convergence -rate classical algorithms as an improvement operator for the elite of SA,which greatly improves SA's ability of local exploitation and accelerates its convergence rate.The numerical simulation results show that the SA-based hybrid approach is fairly robust to initial conditions and has high convergence rate and solutions precise,and its performance is fairly superior to that of single methods.The algorithms design ideas applied in this paper have great use for reference significance for solving practical problems.
出处
《计算机工程与应用》
CSCD
北大核心
2004年第23期36-39,共4页
Computer Engineering and Applications
基金
国家863高技术研究发展计划课题(编号:2002AA001006)
关键词
高维复杂函数
混合全局优化
模拟退火算法
NFL定理
精英策略
complex functions with high-dimension,hybrid global optimization,Simulated Annealing(SA),NFL theorem,elite approach
