Universal approximation theorem: Difference between revisions

Revision as of 15:56, 23 May 2011

Revision as of 16:03, 23 May 2011

Line 1:

In mathematics, the '''universal approximation theorem''' states<ref>Balázs Csanád Csáji. Approximation with Artificial Neural Networks; Faculty of Sciences; Eötvös Loránd University, Hungary</ref> that the standard [[Multilayer_perceptron|multilayer]] [[feedforward neural network|feed-forward]] network with a single hidden layer that contains finite number of hidden [[neuron]]s, and with arbitrary activation function are universal approximators on a compact subset of [[Euclidean space|'''R'''ⁿ]].

The [[theorem]] was first proved{{cnCitation needed|date=May 2011}} by [[George Cybenko]] in 1989 for a [[sigmoid function|sigmoid]] activation function, thus it is also called the '''Cybenko theorem'''.<ref name=cyb>Cybenko., G. (1989) [http://actcomm.dartmouth.edu/gvc/papers/approx_by_superposition.pdf "Approximations by superpositions of sigmoidal functions"], ''Mathematics of Control, Signals, and Systems'', 2 (4), 303-314</ref>

Kurt Hornik (1991){{cnCitation needed|date=May 2011}} showed that it is not the specific choice of the activation function, but rather the multilayer feedforward architecture itself which gives neural networks the potential of being universal approximators. The output units are always assumed to be linear. For notational convenience we shall explicitly formulate our results only for the case where there is only one output unit. (The general case can easily be deduced from the simple case.)

The theorem<ref name=cyb/><ref>