There exists a neural network that does not make avoidable mistakes

By A. Gallant, Halbert White. 1988

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Implicit Neural Representations with Periodic Activation Functions (2020, background): “Early work includes Fourier neural networks, engineered to mimic the Fourier transform via single-hiddenlayer networks [19, 20].”
Linear, non-linear and essential foreign exchange rate prediction with simple technical trading rules (1999, background): “Many authors have investigated the universal approximation properties of neural networks (Gallant and White, 1988, 1992; Cybenko, 1989; Funanhashi, 1989; Hecht-Nielsen, 1989; Hornik et al., 1989, 1990).”
Original Contribution: Multilayer feedforward networks with a nonpolynomial activation function can approximate any function (1993, unknown): “For example, Gallant and White ( 1988 ) proved that a network with "cosine squasher" activation functions possess all the approximations properties of Fourier series representations.”
Multilayer feedforward networks are universal approximators (1989, background): “Useful examples of squashing functions are the threshold functions, Y(n) = lljpO1 (where l,., denotes the indicator function), the ramp function, *(;l) = ~I[O~-l~lj + 111>1}> and the cosine squasher of Gallant and White (1988), q(J) = (1 -+ cos[J. + 3~121) (li2) l{ -n/Zci~ni2] + l{l>nO).”
Chapter 76 Large Sample Sieve Estimation of Semi-Nonparametric Models (2007, background): “See e.g. Amemiya (1985, chapter 4), Gallant and White (1988b), Newey and McFadden (1994) and Gourieroux and Monfort (1995).”
Extreme learning machines: a survey (2011, background): “Such activation functions include the threshold, ramp and sigmoid functions as well as the radial basis, ‘‘cosine squasher’’ [14] and many nonregular functions.”