Identification of Linear and Nonlinear Sensory Processing Circuits from Spiking Neuron Data

摘要

Inferring mathematical models of sensory processing systems directlyrnfrom input-output observations, while making the fewest assumptionsrnabout the model equations and the types of measurements available, isrnstill a major issue in computational neuroscience. This letter introducesrntwo new approaches for identifying sensory circuit models consisting ofrnlinear and nonlinear filters in series with spiking neuron models, basedrnonly on the sampled analog input to the filter and the recorded spikerntrain output of the spiking neuron. For an ideal integrate-and-fire neuronrnmodel, the first algorithm can identify the spiking neuron parameters asrnwell as the structure and parameters of an arbitrary nonlinear filter connectedrnto it. The second algorithm can identify the parameters of the morerngeneral leaky integrate-and-fire spiking neuron model, as well as the parametersrnof an arbitrary linear filter connected to it. Numerical studiesrninvolving simulated and real experimental recordings are used to demonstraternthe applicability and evaluate the performance of the proposedrnalgorithms.