基于分布式并行计算的神经网络算法

为了提高计算性能（速度与可扩展性），提出了一种新颖的神经网络的并行计算体系结构和计算网络权函数的训练算法。权函数是广义Chebyshev多项式和线性函数的复合函数，只需要通过代数计算就可以求得，不需要梯度下降计算或者矩阵计算。各个权函数能够独立求解，可以通过并行系统采用并行算法计算。算法可以求得全局最优点，得到反映网络误差的一个有用的表达式。此外,算法在不超过权函数总数的范围内，还具有维持加速比与并行系统中提供的处理器的数量成线性增长的能力。仿真实验结果表明，本文算法的计算性能远远优于传统算法。

Training algorithm for neural networks based on distributed parallel calculation

To improve computing performance(speed and scalability),an innovative parallel computation architecture and a training algorithm for neural networks are proposed.Each weight function is a composite function of a generalized Chebyshev polynomial and a linear function,only algebraic calculation is needed,and no requirement is involved for the calculation such as the steepest descent-like algorithms or matrix calculation.The weight functions are found independently,therefore they are calculated by using a para...