Research On Non-Iterative Forword Neural Network Model

Human beings have a highly developed brain, which is the material basis of the thinking activity, and thinking is a concentrated expression of the human intelligence. Artificial neural network is designed to simulate the human brain structure and function of the brain intelligent information processing system, with its unique parallel processing, distributed storage, adaptive self-learning and a high degree of fault tolerance and self-organizing more and more academic widespread concern in industry and business, has been widely applied to the modernization of the national economy and national defense science and technology is a hot research in recent years one of the areas.BP neural network is the most mature and most widely used artificial neural network model, but the network model of slow convergence and easily trapped into local minima, is difficult to determine the network structure and network generalization can not guarantee the inherent weaknesses of, which greatly restricted the application of BP neural network in the field of real-time calculation, automatic control and other demanding real-time.In this paper, design a non-iterative forward neural network model to solve the above problems and shortcomings of the existing BP neural network. Nonlinear mathematical theory based on artificial neural network to explore the basic principles of the BP neural network and defects of focus from the network architecture, activation function and learning algorithm of BP neural network, a new optimized design. Finally, the development prior to the neural network model in the non-iterative, proved the proposed network model with fast learning speed, high prediction accuracy and noise suppression ability and comprehensive advantages to avoid the problem of the iterative computation and local minimum...