基于集群式系统的GSAD算法

针对新出现的高性能价格比的集群式计算方式 ,提出了设计高效 SPMD(single program multiple data)算法的几个原则 ,并基于这些原则 ,给出了求解多极值点优化问题的 GSAD(genetic sim ulated annealing and downhill)算法的描述。该算法有机地结合了遗传算法、模拟退火以及下山的优点 ,达到了高效、收敛、可扩展的效果。基于 MPI编程实现 ,给出了该算法在几个典型的多极值点函数以及实际问题中的应用效果 ,通过与相关工作的简单对比指出了该算法的适用范围和特色。建立 SPMD求解模型是 SPMD算法深入研究的方向

GSAD algorithm for cluster system

Cluster computing is a cost effective method to implement parallel computing and the single program multiple data (SPMD) algorithm of cluster computing can greatly improve the performance of cluster applications. Several principles for the design of high performance SPMD cluster algorithms were used to develop a genetic simulated annealing and downhill (GSAD) SPMD algorithm. The algorithm provides good performance, convergence and scalability by combining features of the genetic algorithm, the simulated annealing algorithm and the downhill algorithm. The GSAD algorithm can resolve various kinds of multiple extrema optimization problems or even NP problems. Test results are given for implementation of the algorithm in MPI for five multiple extrema functions and two real applications. The test examples show the algorithm's high performance relative to related algorithms.