Research on grouping-cascaded BP network model

To resolve the training problem of high dimension BP neural network with limited small samples, this paper puts forward the concept of loosely and tightly grouping-cascaded BP network model, the definition of equivalence with BP neural network, and relative theorem. On the base of constructing the grouping-cascaded model which is proved equivalent to BP network, the required training sample numbers of two kinds of neural network models are compared. Finally, the feasibility and validity of the proposed grouping-cascaded BP network model are verified with simulation results.