A Spiking Neuron and Population Model Based on the Growth Transform Dynamical System.

Area of Science:

• Neuromorphic Engineering
• Computational Neuroscience
• Artificial Intelligence

Background:

• Traditional neuromorphic models use bottom-up or top-down approaches for neural population modeling.
• Existing methods often link energy functionals to statistical measures like firing rates, limiting control over neuro-dynamical parameters.
• A need exists for models allowing independent optimization of neuronal and network dynamics.

Purpose of the Study:

• Introduce a new spiking neuron and population model derived from network energy functionals.
• Enable independent control over steady-state population dynamics, action potential shape, and spiking statistics.
• Demonstrate the model's application in creating efficient spiking associative memories.

Main Methods:

• Developed a novel spiking neuron and population model based on Growth Transform dynamical systems.
• Derived neuronal and population dynamics directly from a network energy functional of continuous-valued neural variables.
• Configured the model to achieve independent control over neuro-dynamical properties.

Main Results:

• Achieved independent control over steady-state population dynamics, action potential shape, and spiking statistics.
• Demonstrated stable and interpretable population dynamics irrespective of network size and connectivity type.
• Constructed a spiking associative memory with reduced spike count and high recall accuracy.

Conclusions:

• The proposed model offers a flexible framework for neuromorphic engineering, allowing precise control over neural dynamics.
• This approach facilitates the development of more energy-efficient and performant neuromorphic systems.
• The model shows promise for applications like associative memory with enhanced efficiency and accuracy.