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

Area of Science:

• Computational Neuroscience
• Systems Neuroscience
• Mathematical Modeling

Background:

• Inferring mathematical models of sensory processing systems from input-output data with minimal assumptions remains a challenge.
• Existing methods often require specific knowledge of model equations or measurement types.

Purpose of the Study:

• To introduce two novel algorithms for identifying sensory circuit models.
• To enable model inference based solely on analog input and spike train output.
• To reduce assumptions about model equations and available measurements.

Main Methods:

• Developed two distinct algorithms for model identification.
• The first algorithm targets nonlinear filters with ideal integrate-and-fire neuron models.
• The second algorithm addresses linear filters with leaky integrate-and-fire neuron models.

Main Results:

• Successfully identified spiking neuron parameters and arbitrary nonlinear filter characteristics.
• Accurately determined parameters for leaky integrate-and-fire neuron models and linear filters.
• Validated algorithms using both simulated and real experimental data.

Conclusions:

• The proposed algorithms offer robust methods for inferring sensory circuit models.
• These approaches advance the field of computational neuroscience by simplifying model identification.
• Demonstrated applicability and performance in diverse numerical studies.