Fault Models for Neural Hardware
Amit Prakash Singh, Pravin Chandra, Chandra Sekhar Rai · 2009
Artificial Neural Networks are inherently fault tolerant. Fault tolerance property of artificial neural networks has been investigated with reference to the hardware model of artificial neural networks. In this paper, we propose a framework for the investigation of fault tolerance properties of a hardware model of artificial neural networks. The result obtained indicates that networks obtained by training them with the resilient back propagation algorithm are not fault tolerant: more experimentation is required before a definitive statement can be made for other training algorithms, like the adaptive learning rate algorithm, the conjugate gradient based training algorithms, etc.