ABSTRACT
An Nth order asymptotic expansion is given for the error of weak approximation of a special class of functions by the squashing neural network operators. This class consists of functions / that are N times continuously differentiable over R, so that all / , / ' , . . . , / ^ have the same compact support and f(N>> is of bounded variation. This asymptotic expansion contains products of integrals of the network activation squashing function S and / . The rate of the convergence depends only on the first derivative of the involved functions. This treatment is based on [21].
