Distributing back propagation networks over the Intel iPSC/860 hypercube

The authors describe a parallel implementation of the backpropagation learning algorithm designed for a hypercube, a general-purpose parallel architecture. By decomposing the backpropagation network, and distributing it over the available processors, the time necessary to perform large simulations was decreased. NETtalk, a network that transforms text to speech, was used to test the performance of the simulator. The implementation was successful. As the number of processors increased, the time required to complete the simulation decreased