A novelty detector using a network of integrate and fire neurons
Information in the nervous system has often been considered as being represented by simultaneous discharges of a large set of neurons. We propose a learning mechanism for neural information processing in a simulated cortex model. Also, a new paradigm for pattern recognition by oscillatory neural networks is proposed. The relaxation time of the oscillatory networks is used as a criterion for novelty detection.
Unable to display preview. Download preview PDF.
- 1.Dayhoff J.E. et al.: “Developing Multiple Attractors in a Recurrent Neural Networks”, Proc. of WCCN'94, San Diego, Jun. 1994, Vol. 4, pp. 710–715.Google Scholar
- 2.Yukio Hayashi: “Oscillatory Neural Networks and Learning of Continuously Transformed Patterns”, Neural Networks, 1994, Vol. 7, No 2, pp. 219–231.Google Scholar
- 3.Hill S., Villa A.: “Global Spatiotemporal Activity Influenced by Local Kinetics in a Simulated «Cortical” Neural Network, Workshop on Supercomputing in Brain Research: from topography to neural networks, 1995, World Scientific, pp. 371–375.Google Scholar
- 4.Matsuno Tet al.: “Periodic Signal Learning and Recognition in Coupled Oscillators”, Journal of Physical Society of Japan, Vol. 63, No. 3, March, 1994, pp. 1194–1204.Google Scholar
- 5.Stassinopoulos D., Bak P.: “Self-organization in a Simple Brain Model”, Proc. of WCNN'94, San Diego, Jun, 1994, Vol. 1, pp. 4–26.Google Scholar
- 6.Thiran P., Hasler M: “Information storage using stable and unstable oscillations: an overview”, Int. Journal of Circuit Theory and Applications, Vol. 24, 57–67, 1996.Google Scholar