船橋賢一
On the Approximate Realization of Continuous Mappings
by Neural Networks
Abstract:In this paper, we prove that any continuous mapping can be
approximately realized by Rumelhart-Hinton-Williams' four-layer
neural networks whose output functions for hidden units are
sigmoid functions. We also show that for the approximate
realization of continuous mapping, output functions need not
always be sigmoid but can also be the sigmoid-like functions
defined in this paper. This fact is proved by applying a lemma to
the Kolmogorov-Arnol'd-Sprecher theorem.