Fault-tolerant back-propagation model and its generalization ability
Yuhao Tan, Takashi Nanya · 2005
This paper presents a learning algorithm for multilayer neural networks that brings out the potential ability of fault-tolerance in the network. Experimental results show that fault-tolerant networks obtained by the proposed algorithm also have better generalization ability. The close relationship between fault-tolerance and generalization ability is discussed with some simulation results that clearly illustrate this property.