Neural dynamics leading to optimized information transfer

Abstract

Neural information processing by trains of action potentials is studied in the context of weakly electric fish electroreceptor neurons. A simple but accurate dynamical model for the firing activity of these neurons is presented and compared with experimental results. Dynamical analysis of the model reveals the mechanism by which it reproduces features present in experimental data, such as relative refractoriness and bursting behaviour. Approximations necessary for application of information theory to neural spike trains are presented and different measures are compared. Finally, the consequences of spike patterning caused by relative refractoriness and bursting on information transfer are investigated. It is found that relative refractoriness can increase information transfer while bursting provides a non-linear mechanism for encoding information that might be more efficient than firing of isolated spikes.