Neuromorphic Analog Implementation of Neural Engineering Framework-Inspired Spiking Neuron for High-Dimensional...

Area of Science:

• Neuromorphic Engineering
• Artificial Intelligence Hardware
• Computational Neuroscience

Background:

• Artificial intelligence (AI) demands high-performance, energy-efficient hardware.
• Brain-inspired computing utilizes neural principles for advanced AI frameworks.
• The Neural Engineering Framework (NEF) provides a theoretical basis for large-scale neural networks using spiking neurons.

Purpose of the Study:

• To present OZ, a programmable analog hardware implementation inspired by the Neural Engineering Framework (NEF).
• To demonstrate dynamic programming of OZ neurons for high-dimensional data representation.
• To achieve efficient and low-power neuromorphic data processing.

Main Methods:

• Developed OZ, a programmable analog circuit implementing NEF-spiking neurons.
• Configured OZ neurons with dynamic, high-dimensional response curves and adjustable encoders.
• Validated hardware correspondence with NEF principles, including firing rates and encoding vectors.

Main Results:

• OZ neurons exhibit full correspondence with NEF across key parameters.
• Independent real-time configuration of OZ neurons enables efficient representation spanning.
• The hardware achieves neuromorphic distributed representation with reduced neuron count and power usage.

Conclusions:

• OZ provides a practical, efficient analog hardware solution for NEF-based neuromorphic computing.
• Programmable analog neurons offer a pathway to scalable and power-efficient AI systems.
• This work advances brain-inspired hardware for sophisticated AI applications.