Capabilities of a three layer feedforward neural network

摘要

Mapping capabilities of a three-layer feedforward neural networknwith a finite number of hidden units which have sigmoid functions asntheir nonlinearities are discussed. It is proved that sigmoid functionsnof a hidden layer of the network can raise the dimension of the inputnspace up to the number of the hidden units. From this result, it isnconcluded that a three-layer feedforward neural network with Nnhidden units can assign arbitrary analog values to N arbitraryninput vectors