Minimization of Error Functionals over Perceptron Networks
Abstrakt
Conditions on the data, which guarantee that a good approximation of global minima of error functionals can be achieved using perceptron networks with a limited complexity, are derived.