An Analytically Transparent Network for Sequence Recognition
In this article technical details of a neural network for sequence recognition are presented. The network is powerful enough to simulate finite-state acceptors, while its analysis is much simplified, compared to the standard Hopfield model. Also its processing speed is optimal, and the presence of mixture states is externally controllable. The network is robust under synaptic noise, it has error correcting properties and it can recover from gross input errors.
KeywordsInput Image External Input Sequence Recognition Temporal Image Mixture State
Unable to display preview. Download preview PDF.
- M.F.J. Drossaers. In Proceedings of COLING’ 92, pages 113–119, 1992.Google Scholar
- J.J. Hopfield. Proc. Natl. Acad. Sci. USA, 79: 2554–2558, 1982.Google Scholar
- D. Kleinfeld. Proc. Natl. Acad. Sci. USA, 83: 9469–9473, 1986.Google Scholar
- D. Servan-Schreiber, A. Cleeremans, and J.L. McClelland. In D.S. Touretzky, editor, Advances in Neural Information Processing systems I, Los Altos (Cal.), 1989. Morgan Kaufmann.Google Scholar