Title :
Some queries on “Comments on `Approximation capability in C(R¯n) by multilayer feedforward networks and related problems´”
Author_Institution :
Dept. of Math., Fudan Univ., Shanghai, China
fDate :
9/1/2001 12:00:00 AM
Abstract :
In the comments letter by Huang et al. (ibid. vol.9 (1998)) the authors claimed that the boundedness of a sigmoidal function σ(x) is neither sufficient nor necessary condition for the validity of the approximation theorem discussed in our original paper (ibid. vol.6 (1995)). In this paper, we show that this claim is incorrect
Keywords :
feedforward neural nets; function approximation; boundedness; feedforward neural networks; function approximation; multilayer neural networks; necessary condition; sigmoidal function; sufficient condition; Equations; Network address translation; Neural networks; Neurons; Nonhomogeneous media; Quantization; Stability; State-space methods; Statistics; Testing;
Journal_Title :
Neural Networks, IEEE Transactions on
