CODING EFFICIENCY AND INFORMATION RATES IN SENSORY NEURONS

Many sensory neurons code information about the sensory world in discrete spike trains. How much information is carried in sensory spike trains and how close are the information rates to fundamental limits imposed by the spiking statistics? By learning to estimate time-dependent sensory input signals from the spike trains of single neurons, we are able to measure the rate at which the spike trains convey information about the sensory inputs. These information rates are within a factor of two of the absolute information-theoretic bounds imposed by the entropies of the spike trains themselves.