Fault tolerance of the backpropagation neural network trained on noisy inputs

Preliminary results of a study to determine the effect of noisy training sets on fault tolerance are presented. Backpropagation was used to train three networks on 7*7 numeral patterns. One network was the control and used noiseless inputs and the other two used two different noisy cases. After learning was complete, the networks were tested for their fault tolerance to stuck-at-1 and stuck-at-0 element faults, as well as weight connection faults. The networks trained on noisy inputs had substantially better fault tolerance than the network trained on noiseless inputs.<<ETX>>