Variable structure and variable learning rate Fourier neural networks research

摘要

On the base of the Fourier neural networks, this paper adopted dichotomy to search the neural networks' optimization structure and optimization learning rate. Given the variational ranges of the Fourier neural networks' structure and learning rate, on the condition of arbitrary nonlinear mapping relationship, arbitrary error request and arbitrary training sample number, this algorithm can adjust the fourier neural networks' structure and learning rate automatically to the optimization structure and the optimization learning rate. The simulation results showed that the convergence speed of the fourier neural networks can be greatly improved if the fourier neural networks adopt the optimization structure and the optimization learning rate.