Population coding in the retina: from data to spin-glass models and optimal codes

Abstract:

In most areas of the brain, information is encoded in the correlated activity of large populations of neurons. Here we ask two questions: (i) How can the architecture of such codes be teased apart from experimental data that records simultaneous activity of many neurons, and what are the qualitative features of these real population codes? (ii) How should neural responses be coupled to best represent information about different ensembles of correlated stimuli, and what are the qualitative features of such optimal population codes? In this comparison of real vs optimal systems we find interesting connections to ideas about spin-glasses and information theory, and are able to predict phenomena like 'network adaptation' that can be tested in the upcoming generation of experiments

back