Back to Departmental Colloquium: Spring 2000
Departmental Colloquium
Date: Thursday, Feb 10, 2000
Time: 4:15PM
Location: JWB 335
Duane Nykamp
Courant Insitute
Title |
A population density approach that facilitates large-scale modeling of neural networks |
Abstract |
The neural networks of even small functional units in the brain are enormously complex. Conventional simulation methods, where one models thousands of individual neurons, can take large amounts of computer time even for models of small cortical areas. The population density approach can be used to speed up large-scale neural network simulations. In this method, one groups neurons into large populations of similar neurons. By calculating the evolution of a probability density function for each population, one obtains population firing rates and the distribution of neurons over state space. I demonstrate a population density method for simulating networks of integrate-and-fire neurons with instantaneous synapses or with slow inhibitory synapses. Through comparisons with conventional Monte-Carlo simulations for a model of a hypercolumn in cat visual cortex, I demonstrate the speed and accuracy of the population density method. |
Probability
Mathematical Biology