In this paper, I described the principle and structure on neural networks. In order to approve the convergence and generalization ability, I brought forward some new methods to train neural networks. After I lucubrated dynamic neural network, a new style dynamic regressive neural network was devised and gained a good result in convergence and generalization ability. I gave the method on how to device a neural network to identify a model or reverse model. In the last chapter, I introduced the knowledge on neural network control.
|