Estimates of Network Complexity and Integral Representations
Abstrakt
Upper bounds on rates of approximation by neural networks are derived for functions representable as integrals in the form of networks with infinitely many units. The bounds are applied to perceptron networks.