基于网格服务的GEP分布式函数挖掘算法

提出了一种基于网格服务的GEP分布式函数挖掘算法(DFMGEP-GS),它将网格服务与GEP算法相结合,既成功地实现了在网格平台下的GEP函数挖掘,又提高了每个网格节点上GEP算法的全局寻优性;同时证明了在网格环境下由局部数据模型生成全局数据模型的方法.仿真实验结果表明,对于函数类型已知的数据,随着数据集的增大,在成功挖掘到目标函数的情况下,DFMGEP-GS算法的平均耗时最少,而且随着网格节点的增加,DFMGEP-GS的收敛速度最大提高了约17倍;对于函数类型未知的复杂数据集,DFMGEP-GS算法挖掘所得到的模型的误差最小.

Distributed Function Mining for GEP on Grid Services

This paper presented distributed function mining for GEP on grid services(DFMGEP-GS),which combined grid services and GEP algorithm to realize not only function finding for GEP on grid platform successfully, improve but also global optimization of GEP algorithm with every grid node.Meanwhile, it proved the method by which global data model is obtained by means of local data model on grid.By simulation experiment,it is showed that for data with known function type, and with the augmentation of datasets, average consumptive time of DFMGEP-GS is less than other three algorithms under the condition of mining target function successfully, and that with the increment of grid nodes, the convergent speed of DFMGEP-GS is improved about 17 times maximally.For very complex data with unknown function type, the error which is mined by DFMGEP-GS algorithm is minimum.