SIMO Fourier neural networks research
Xu-Hua Yang, Huaping Dai, Youxian Sun · 2004
This paper proposed the single input multiple outputs (SIMO) Fourier neural networks on the base of Fourier series principle. The SIMO Fourier neural networks turn nonlinear optimization problem into linear optimization problem. So, the SIMO Fourier neural networks highly improve convergence speed and avoid local minima problem. At the same time, under the condition of bounded input and bounded output, the SIMO Fourier neural networks can approximate multiple arbitrary nonlinear mapping relationship at arbitrary accuracy and have good generalization capability.