On the possible use of ICA to identify synaptic inputs from observations of several neurons

Pando Georgiev and Henry Tuckwell

Contact the author: Please use for correspondence this email.
Submission date: 03. Dec. 2004
published in: Neurocomputing, 67 (2005), p. 450-455 
DOI number (of the published article): 10.1016/j.neucom.2004.08.008
Bibtex
Keywords and phrases: ica for de's, synaptic input, neuron

Abstract:
We consider the problem of separating and determining the time courses of various synaptic input currents from simultaneous recordings of the time courses of membrane potentials, including spikes, of several neurons. Employing a suitable mathematical model the method involves the differentiation of potentials and the use of ICA to determine the relative strengths of various synaptic inputs. At the same time the waveforms of those input currents are recovered. We illustrate the application to nonlinear point models with deterministic and stochasticinput currents,using a single-component Fitzhugh-Nagumo approximation.