Study and Implementation of BP Network Based on Serial-parallel Computing Topology

The topology of an artificial neural network (ANN), characterized by connectionism, is an important factor which affects the function and the performance of ANN. In this paper, based on analysis of some existing neural network models, a method for constructing BP networks using serial-parallel computing is proposed. "Ordering", one of the three sample feature problems, is solved topologically. Analysis shows that precision and learning speed of ANN are enhance remarkably.