Fault tolerant neural networks in optimization problems
Y. Koyanagi, Yoshihiro Tohma · 2003
The authors discuss the influence of stuck-at faults in neural networks for solving optimization problems. They use a Hopfield model of a neural network, applying it to the traveling salesman problem of five cities. The asymmetric nature of fault tolerance of the network against stuck-at-zero and stuck-at-one faults is revealed. A method to alleviate this asymmetry and enhance the fault tolerance greatly is proposed.>