Research Of Neural Network And Adaptive Control Using Neural Network

As a new intelligent control method, the neural-network-based technique provides a potential tool for solving the control problems of highly uncertain, complex, nonlinear dynamical systems. This paper is mainly concerned with learning algorithms of feedforward neural networks, and adaptive control of nonlinear discrete time dynamical systems using neural networks. The main contributions of this paper are as follows:1.The influence of activation functions in BP algorithms is studied, and trying to amend it.2.In order to improve the convergence characteristics, we discuss the strong effect of the initial weight to the training speed based on the theory of the BP algorithm, which is proved by the results.3.We make use of emendation algorithm. This is effective to expedite training and avoid local area.4.Combining adaptive control and neural network, we establish two models, linear robust adaptive controller and neural networks. The order of improving performance is owned.