For many problems in classification compensation, adaptivity,identification, and signal processing, results concerning therepresentation and approximation of nonlinear functions can be ofparticular interest to engineers. Here we consider a large class offunctions f that map R' into the set of real or complex numbers, andwe give bounds on the number of parameters of a certain approximationnetwork so that f can be approximated to within a prescribed degreeof accuracy using an appropriate configuration of the network. Wealso describe related work in the neural networks literature.
展开▼
