摘要
为了加快在大规模神经网络训练下并行技术的训练速度问题,从BP算法的内部结构分析了BP神经网络算法的大规模行划分方法,提出了一种动态负载平衡方案。通过在PC集群环境下对并行算法的试验结果表明,这种并行划分提高了加速比,具有现实意义。
In this paper, a parallel BP neural network algorithm is presented in order to speed up the training of large scale neural networks. And presents a dynamic load balancing scheme. To demonstrate the gain in performance provided, this algorithm is realized in MPI programming environment of the PC cluster. The result indicates that the main goal of speeding up the computation was achieved.
出处
《计算机技术与发展》
2006年第7期67-69,72,共4页
Computer Technology and Development
关键词
BP神经网络
并行
动态负载平衡
BP neural network
parallel
dynamic load balancing
