A NOVEL THEOREM ON THE MULTI-DIMENSIONAL FUNCTION APPROXIMATION ABILITY OF FEED FORWARD MULTI-LAYER NEURAL NETWORKS

韦岗,贺前华.神经网络模型学习及应用,北京:电子工业出版社,1994，第三章.[2]Sam Kwong, Wei Gang, Ouyang Jing-zheng. Discrete utterance recognition based on nonlinear[3]model identification with single layer neural networks. Proc. IEEE Int. Conf. Circuits and Systems, USA:1993, 2419-2422.[4]Wei Gang, Ouyang Jing-zheng. On the bound of the approximation capacity of multi-layer neural[5]networks. Proc. Int. Joint. Conf. Neural Networks, Singapore:1991, 2299-2304.[6]Cybenko G. Approximation勿superposition of a sigmoidal function[J].Math. Control, Signals, Syst.1989, 2(4):303-314[7]Hartman E J, Keeler J D, Kowalski J M. Layered neural networks with Gaussian hidden units as universal approximations[J].Neural Comp.1991, 2(3):210-215[8]Hornik K. Approximation capacities of multilayer feedforward networks[J].Neural Networks.1991, 4(3):251-257[9]Park J, Sandberg I W. Universal approximation using radial-basis-function networks[J].Neural Comp.1991, 3(3):246-257[10]Park J, Sanberg I W. Approximation and radial-basis-function networks[J].Neural Comp.1993, 5(4):305-316