Fault tolerant constructive algorithm for feedforward neural networks

In this paper, a constructive algorithm for fault tolerant feedforward neural network, called FTCA, is proposed. The algorithm starts with a network with a single hidden neuron, and a new hidden unit is added to the network whenever it fails to converge. Before inserting the new hidden neuron into the network, only the weights connecting the new hidden neuron to the other neurons are trained (i.e. updated) until there is no significant reduction of the output error. To generate a fault tolerant network, the relevance of synaptic weights is estimated in each cycle. And only the weights which have a relevance less than a specified threshold are updated in that cycle. The loss of connections between neurons (which are equivalent to stuck-at-0 faults) are assumed. The simulation results indicate that the network constructed by FTCA has a significant fault tolerance ability.