A Learning Algorithm with Activation Function Manipulation for Fault Tolerant Neural Networks

摘要

In this paper we propose a learning algorithm to enhance the fault tolerance of feedforward neural networks (NNs for short) by manipulating the gradient of sigmoid activation function of the neuron. We assume stuck-at-0 and stuck-at-1 faults of the connec- tion link. For the output layer, we employ the function with the rela- tively gentle gradient to enhance its fault tolerance. For enhancing the fault tolerance of hidden layer, we steepen the gradient of function af- ter convergence.