Research And Application Of Learning Algorithm Of Neural Network Based Cosine

The research of artificial neural networks is more and more concerned, and a lot of results of its theories, methods with the applications are acquired. The basic model and characteristics of artificial neural networks (ANN, such as BP network, Hopfield network, RBF network etc.) and its developmental direction are briefly introduced in this paper. This project aimed to using cosine-basis neural network algorithms that optimize the design of conventional algorithms. This paper proposed a cosine-basis neural network model. We also analyzed the network structure; of this model and attested a few important theoretical arguments. The new algorithm was then applied in two cases: numerical analysis (linear equations, nonlinear equations, numerical integral etc.) and the optimization of FIR digital filter design. Simulation results showed that the proposed new algorithm outperformed traditional algorithms in terms of both accuracy and converging rate and it proved to be a very practical and effective ANN-based algorithm.