Partially weight minimization approach for fault tolerant multilayer neural networks

We propose a new learning algorithm to enhance fault tolerance of multilayer neural networks (MLNs). This method is based on the fact that strong weights make MLNs sensitive to faults. To decrease the number of strong connections, we introduce a new evaluation function for the new learning algorithm. The function consists of two terms: one is the output error and the other is the square sum of HO-weights (weighs between the hidden layer and output layer). The second term aims to decrease the value of HO-weights. By decreasing the value of only HO-weights, we enhance the fault tolerance against the previous method.