Dynamic spike threshold and nonlinear dendritic computation for coincidence detection in neuromorphic circuits

Area of Science:

• Neuroscience
• Electronic Engineering
• Materials Science

Background:

• Artificial neurons aim to replicate biological neuron functions for advanced computing.
• Understanding neuronal dynamics, including spike thresholds and dendritic activity, is key to developing more sophisticated artificial neural networks.
• Carbon nanotube field-effect transistors (CNFETs) offer potential for creating compact and efficient electronic neuron circuits.

Purpose of the Study:

• To design and simulate an electronic cortical neuron model with dynamic spike threshold and active dendritic properties.
• To investigate the neuron's firing behavior in response to synchronous and asynchronous synaptic inputs.
• To evaluate the role of dendritic spikes in precise input-output transformation and resilience to timing variations.

Main Methods:

• Development of an electronic cortical neuron circuit incorporating dynamic spike threshold and active dendritic properties.
• Simulation of the circuit using a carbon nanotube field-effect transistor (CNFET) SPICE model.
• Analysis of the neuron's response to varying synaptic input timing and synchronization patterns.

Main Results:

• The electronic neuron exhibits a lower spike threshold for coincident synaptic inputs.
• Asynchronous synaptic inputs require greater depolarization to elicit neuronal firing.
• Dendritic spikes were shown to be critical for precise input-output transformation, reliable firing, and enhanced resilience to input jitter.

Conclusions:

• The developed electronic neuron model successfully mimics key aspects of biological neuronal function, including dynamic thresholding and dendritic integration.
• The findings highlight the importance of dendritic spikes for achieving precise and robust neural information processing in artificial systems.
• This work contributes to the advancement of neuromorphic computing by providing a novel electronic neuron design with improved computational capabilities.