Phil Nelson (Biological Physics, University of Pennsylvania, USA)
Abstract:To directly test for adaptive decorrelation in guinea pig retina, we measured responses of simultaneously recorded populations of ~40 retinal ganglion cells presented with three classes of stimuli: spatially uncorrelated, exponentially correlated, and naturalistically correlated (approximately scale-invariant). In the two latter cases, the retinal output was significantly decorrelated relative to the input. Responding to spatially co-varying stimuli, cells had wider receptive fields, and their static nonlinearities had gains that scaled inversely with a measure of `effective contrast', which increased with stimulus correlations. Our measurements were done using a new Bayesian spike sorting algorithm.
JS Prentice et al. Fast, scalable, bayesian spike identification for multi-electrode arrays. PLoS ONE (2011) vol. 6 (7) pp. e19884