Effects of the learning algorithm on the faulty behavior of feedback neural networks
Zhang Tao · Journal of Tsinghua University(Science and Technology) · 2000
This paper studies the faulty behavior of feedback neural networks with stuck at faults during the course of learning using the discrete hopfield neural network (DHNN) as an example. DHNN and its Hebb learning algorithm are briefly introduced. Then the study analyzes the influence of link faults on the weights during learning and gives formulas for output probability distribution verification. A computer simulation verifies the accuracy the theoretical analysis.