Fault-diagnosis of digital circuits using neural network of hybrid learning algorithm
Yong‐Hyun Cho, Yong-Su Park · 1999
This paper proposes a new hybrid learning algorithm for multilayer neural networks and an efficient neural network based diagnostic system for digital circuits. A hybrid learning algorithm is combined to the steepest descent method and dynamic tunneling system. The steepest descent method is applied for high-speed learning, the dynamic tunneling system which has a tunneling phenomenon, for global learning. The proposed fault-diagnosis system has been applied to the parity generator circuit. The simulation results show that the system using the proposed learning algorithm is higher convergence speed and rate, in comparison with system using the conventional backpropagation algorithm.