Modeling networks with linear (VLSI) integrate-and-fire neurons
We analyse in detail the statistical properties of a “canonical” integrate-and-fire neuron with a linear integrator as often used in VLSI implementations . We show that a network of such elements can maintain both stable spontaneous activity and selective (stimulus specific) activity, contrary to current opinion. The spike statistics appears to be qualitatively the same as in networks of conventional (exponential) integrate-and-fire neurons that in turn, exhibit a wide variety of characteristics observed in cortical recordings.
KeywordsInput Current Synaptic Efficacy VLSI Implementation Stable Fixed Point Negative Drift
Unable to display preview. Download preview PDF.
- 1.C. Mead: Analog VLSI and neural system. Reading, MA: Addison Wesley (1989)Google Scholar
- 2.D.J. Amit and N. Brunel: Model of global spontaneous activity and local structured activity during delay periods in the cerebral cortex. Cerebral Cortex, 7, 3 (1997)Google Scholar
- 3.M. Annunziato: Hardware implementation of an attractor neural network with IF neurons and stochastic learning, thesis (in italian) (1995)Google Scholar
- 4.G.L. Gerstein and B. Mandelbrot: Random walk models of the spike activity of a single neuron. Biophysical Journal, 4 (1964) 41–68Google Scholar
- 5.Cox Miller: D.R. Cox and H.D. Miller, The theory of stochastic processes. London, Methuen and Co LTD. (1965)Google Scholar
- 6.C. A. van Vreeswijk and M.E. Hasselmo, Self-sustained memory states in a simple model with excitatory and inhibitory neurons. unpublishedGoogle Scholar
- 7.M. Mattia, Dynamics of an IF linear neurons network, thesis (in italian) (1997)Google Scholar