Contrast Adaptation By Adjusting Neurotransmitter Release Probability

ng, depressing or facilitating synapse. To derive a learning rule we again make the hypothesis that the mutual information between a cortical neurons output ring rate and its geniculate input ring should be maximal. For the output ring rate we use an approximation of the contrast response function of the model. We show that the experimental ndings of contrast adaptation (Carandini & Ferster 1997) can be reproduced by regulating neurotransmitter release at the aerent but not recurrent connections. Speci cally dierent temporal components of the membrane potential and spike response at the optimal and near optimal orientations in our model show qualitatively the same behavior as in the experiments. We also observe a phase advance relative to the driving stimulus with adaptation, that is with reduction in release probability, as has also been experimentally observed. Furthermore our model shows a signi cant decrease in the maximal response as well as a large shift of the contrast