Two-phases Parallel Neural Network Algorithm Based on RPROP

BP algorithm is widely used in the field of business intelligence. Aimed at improving its relatively slow convergence speed and its tendency to be trapped in local minima easily, an improved two-phases parallel algorithm is presented in this paper. The first parallel operation is to cast about for minima area so as to avoid getting into local optimal solution to some extent and accelerate convergence, thereby reducing the number of epochs. The second parallel operation is to shorten learning time. The experiments demonstrate that the improved two-phases parallel BP algorithm has the better performance of speedup.