Competitive learning behavior in a stochastic neural network
- 46 Downloads
Stochastic behavior is a natural and inevitable property of biological neurons. The effect of stochastic behavior or thermal fluctuation in neural firings on the learning process in a neural system is investigated. A learning model, which is derived from the stochastic differential equation of the firing-rate model, is presented as an estimate of the gradient flow of free energy. The model reveals that the learning process becomes competitive owing to the effect of entropy even through the synapse modifications only follow the simple Hebbian rule.
KeywordsStochastic neural network Competitive learning
Unable to display preview. Download preview PDF.
- D. O. Hebb, The organization of behavior: A neurophysiological theory (John Wiley and Sons, New York, 1949).Google Scholar
- E. L. Bienenstock, L. N. Cooper and P. W. Munro, J. Neurosci. 2, 32 (1982).Google Scholar
- G.-q. Bi and M.-m. Poo, J. Neurosci 18, 10464 (1998).Google Scholar
- L. F. Abbott and S. B. Nelson, Nat. Neurosci 3, 1173 (2000).Google Scholar
- C. Peipenbrock, in Probabilistic Models of the Brain, edited by R. P. N. Rao, B. A. Olshausen and M. S. Lewicki (The MIT Press, Cambridge MA, 2002), p. 181.Google Scholar