Gauss-Chebyshev neural networks
Hong-Jie Xing, Bao-Gang Hu · 2005
This paper presents a novel neural network integrating both Gauss neural network and Chebyshev neural network. The Gauss-Chebyshev neural networks take advantages of the for local approximation ability, but the Chebyshev one for global generalization ability. Numerical experiments confirm the new strategy on the better performance in comparison with Gauss neural networks. Furthermore, under the same initialization conditions, Gauss-Chebyshev neural network is more efficient than Gauss-Sigmoid neural network for regression application. All eight functions tested from the experiments show the improvements of the proposed neural networks.