Information Based Limits on Synaptic Growth in Hebbian Models
Hebb’s seminal concept that synaptic strength varies as a function of correlated pre- and post-synaptic activity has impacted pharmacologic, physiologic, and computational models of the nervous system [Hebb, 1949]. Specifically, he proposed that correlated pre- and post-synaptic activity leads to an increase in synaptic strength. When computational models incorporate Hebbian synapses, some limit must be placed on synaptic growth if well-defined solutions are to be produced, and it may be necessary to incorporate some means by which synaptic strength can also decrease. Strategies to limit synaptic growth have included some type of limit on individual or total synaptic size. Another approach incorporates a computationally [Bienenstock, Cooper, and Munro, 1982] motivated “sliding threshold” for synaptic modification where the sliding threshold is stimulus dependant.
Unable to display preview. Download preview PDF.
- Albrecht. D.G., and Hamilton D.B. Striate cortex of monkey and cat: Contrast response function. J. Neurophysiol. 48: 217 237. 1982.Google Scholar
- Gottschalk, A., McLean, J., Palmer, L.A. Contrast response in the context of a cortical model which maximally preserves information transfer to higher visual centers. Invest. Opthal. Vis. Sci. 34: 909. 1993.Google Scholar
- Gottschalk. A. Limits on image representation in early vision. In: Computation in Neurons and Neural Systems. Eeckman, F.H. (ed.). Boston: Kluwer Academic Publishers, 1994.Google Scholar
- Gottschalk, A. Maximizing mutual information in a simple model of early vision. Submitted for publication. 1996.Google Scholar
- Hebb, D.O. The Organisation of Behavior. New York, John Wiley, 1949.Google Scholar
- Ohzawa, I., Sclar. G.. and Freeman, R.D. Contrast Bain control in the cat’s visual cortex.,I. Neurophysiol. 54: 651–667. 1985.Google Scholar