The storage capacity of multilayer networks with overlapping receptive fields is investigated for a constructive algorithm within a one-step replica symmetry breaking (RSB) treatment. We find that the storage capacity increases logarithmically with the number of hidden units <span class='mathrm'>K</span> without saturating the Mitchison-Durbin bound. The slope of the logarithmic increase decays exponentionally with the stability with which the patterns have been stored.
|Title of host publication||Mathematics of Neural Networks: Models, Algorithms and Applications|
|Editors||Stephen W. Ellacott, John C. Mason, Iain J. Anderson|
|Place of Publication||Oxford|
|Number of pages||7|
|Publication status||Published - 1997|
Bibliographical noteThe original publication is available at www.springerlink.com
- human factors