Enhancing Retrieval in Neural Networks
M. Andrecut · Modern Physics Letters B · 1998
A simple method for retrieving information in Hopfield neural network is discussed. Numerical simulations for different network sizes were performed in order to test the ability of the model in retrieving patterns. It is shown that an imposed external field (stimulus) enhances the retrieval results of the neural network. For random and unbiased patterns the critical saturation parameter is enhanced from α=0.14 up to α=0.5. Also, important enhancements of the critical saturation parameter are obtained low-activity patterns.