A Low-Cost Fault-Tolerant Approach for Hardware Implementation of Artificial Neural Networks

Artificial Neural Networks (ANNs) are widely used in computational and industrial applications. As technology is developed the scale of hardware is progressively becoming smaller and the number of faults is increasing. Therefore, fault-tolerant methods are necessary especially for ANNs used in critical applications. In this work, we propose a new method for fault-tolerant implementation of neural networks. In hidden and output layers, we add a spare neuron, and one of hidden and output neurons is tested by each input pattern. Our technique detects and corrects any single fault in the network. We achieve complete fault tolerance for single faults with at most 40% area overhead.

[1]  J. I. Minnix.  Fault tolerance of the backpropagation neural network trained on noisy inputs , 1992, [Proceedings 1992] IJCNN International Joint Conference on Neural Networks.

[2]  Chilukuri K. Mohan,et al.  Modifying training algorithms for improved fault tolerance , 1994, Proceedings of 1994 IEEE International Conference on Neural Networks (ICNN'94).

[3]  Itsuo Takanami,et al.  A Multiple-Weight-and-Neuron-Fault Tolerant Digital Multilayer Neural Network , 2006, 2006 21st IEEE International Symposium on Defect and Fault Tolerance in VLSI Systems.

[4]  A Learning Algorithm for Fault Tolerant Feedforward Neural Networks , 1996 .

[5]  Dhananjay S. Phatak,et al.  Complete and partial fault tolerance of feedforward neural nets , 1995, IEEE Trans. Neural Networks.

[6]  Lutz Prechelt,et al.  A Set of Neural Network Benchmark Problems and Benchmarking Rules , 1994 .

[7]  Itsuo Takanami,et al.  Learning Algorithms Which Make Multilayer Neural Networks Multiple-Weight-and-Neuron-Fault Tolerant , 2008, IEICE Trans. Inf. Syst..

[8]  Robert Chun,et al.  Immunization of neural networks against hardware faults , 1990, IEEE International Symposium on Circuits and Systems.