Modelling Predictive Information of Stochastic Dynamics in the Retina

Many experiments showed that the retina processes information before transmitting them to the visual cortex. We propose a model to elucidate the predictive effect of the amacrine cells and ganglion cells in the retina. We generate the input signals with OU (Ornstein-Uhlenbeck) and HMM (Hidden Markov model) process, and compare the mutual information calculated from the simulations with those of the moving bar experiments on bullfrog retina mounted on a multi-electrode array, illustrating that the model agrees with the experiments.