函数优化问题的一种异步并行模拟退火算法  

An asynchronous parallel simulated annealing algorithm for function optimization problems

作　　者：王华[1] 唐国金[1] 

机构地区：[1]国防科技大学航天与材料工程学院,湖南长沙410073

出　　处：《控制与决策》2005年第5期579-582,共4页Control and Decision

摘　　要：针对工程中的多极值点复杂函数的优化问题,提出一种完全异步的粗粒度并行模拟退火算法.在一定时间内将一条Markov链分裂成多条Markov链,并结合其他多种改进方法,获得了可扩展的并行效果,提高了算法应用的灵活性.数值计算表明,该方法可显著提高算法的收敛速度.To complex function optimization problems with multi-minimize in practice, an asynchronous parallel simulated annealing algorithm is (developed,) which splits one Markov chain into several Markov chains, and combines other enhanced method. The (performance) of algorithm scales up with the increase of processors, and the character of asynchronous makes algorithm more suitable for all kinds of processor. The numerical computation results indicate that the proposed algorithm can significantly improve the convergence speed.

关 键 词：模拟退火 并行算法 函数优化 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]