Compact digital implementation of a quadratic integrate-and-fire neuron

Area of Science:

• Computational neuroscience
• Digital hardware implementation
• Neural modeling

Background:

• The quadratic integrate-and-fire (QIF) model is a fundamental computational neuroscience tool.
• Efficient digital implementations are crucial for large-scale neural network simulations.

Purpose of the Study:

• To develop a compact, fixed-point digital implementation of the QIF neural model.
• To derive design equations for optimizing bit representation and hardware resource usage.

Main Methods:

• Derivation of equations for minimum bit requirements for QIF states and switching threshold.
• Minimization of multiplier size for the nonlinear squaring function.
• Development of test vectors to validate all four QIF states.
• Field-programmable gate array (FPGA) implementation.

Main Results:

• A compact fixed-point digital QIF model was successfully implemented.
• Design equations enabled efficient representation of all QIF states.
• The FPGA implementation proved computationally efficient, using only two fixed-point adders and one multiplier.

Conclusions:

• The developed digital QIF model offers a computationally efficient solution for neural simulations.
• The design methodology allows for optimized hardware resource utilization.
• This implementation is suitable for applications requiring compact and efficient neural modeling.