A mapping approach for designing neural sub-nets

Summary form only given. Several investigators have constructed backpropagation (BP) neural networks by assembling smaller, pretrained building blocks. This approach leads to faster training and provides a known topology for the network. The authors have carried this process down one additional level by describing methods for mapping given functions to subblocks. First, polynomial, approximations to the desired function were found. Then the polynomial was mapped to a BP network, using an extension of a constructive proof to universal approximation