On the Convergence Properties of the Temporal Kohonen Map and the Recurrent Self-Organizing Map
This paper compares two Self-Organizing Map (SOM) based approaches for temporal sequence processing: The Recurrent Self-Organizing Map (RSOM) and Temporal Kohonen Map (TKM). The convergence properties of these algorithms are studied, and their difference in learning is emphasized both theoretically and with simulations. The results show that RSOM is superior over TKM.
KeywordsWeight Vector Input Space Input Pattern Optimal Weight Leaky Integrator
Unable to display preview. Download preview PDF.
- M. Varsta, J. Heikkonen, and J. del Ruiz Millán. A recurrent self-organizing map for temporal sequence processing. In Proceedings of the ICANN’97. Springer-Verlag Berlin Heidelberg New York, October 1997. ISBN 3-540-63631-5.Google Scholar
- M. Varsta, J. Heikkonen, and J. del R. Millán. Context learning with the self organizing map. In Proceedings of WSOM’97, Workshop on Self-Organizing Maps, Espoo, Finland, June 4–6, pages 197–202. Helsinki University of Technology, Neural Networks Research Centre, Espoo, Finland, 1997. ISBN 951-22-3589-7.Google Scholar
- T. Kohonen. Self-Organizing Maps, volume 30 of Lecture Notes in Inform. Sciences. Springer, second edition, 1997.Google Scholar
- M. Cottrell. Theoretical aspects of the som algorithm. In Proceedings of WSOM’97, Workshop on S elf-Organizing Maps, Espoo, Finland, June 4–6, pages 246–267. Helsinki University of Technology, Neural Networks Research Centre, Espoo, Finland, 1997.Google Scholar