Basins of Attraction in Neural Network Models Trained with External Fields

The basin of attraction of a neural network model trained by ensembles of external, noisy fields is studied and comparisons made with a statistically identical field applied during retrieval. Introduction of these fields enlarges the basin boundary, with the equal training and retrieval field case having marginally the largest maximum value. However, similarities with the retrieval-field only case suggests this is unlikely to be the optimal relationship.