Computation via Neuron-like Spiking in Percolating Networks of Nanoparticles

Area of Science:

• Computational Neuroscience
• Materials Science
• Artificial Intelligence

Background:

• The biological brain processes information efficiently using electrical spikes.
• Neuromorphic computing aims to replicate brain principles for advanced AI and low-power edge computing.
• Percolating networks of nanoparticles (PNNs) show promise for natural computation due to critical spiking behavior.

Purpose of the Study:

• To demonstrate that PNNs can perform computational tasks using a rate coding scheme.
• To achieve high accuracy in Boolean operations and image classification with PNNs.
• To elucidate the mechanism behind PNN computational capabilities.

Main Methods:

• Utilizing a rate coding scheme with PNNs.
• Manipulating spiking activity via control voltages.
• Analyzing the role of nanoscale tunnel gaps and their nonlinear transformation of input data.

Main Results:

• PNNs successfully performed Boolean operations and image classification tasks.
• Near-perfect accuracy was achieved by controlling spiking activity.
• Nanoscale tunnel gaps were identified as key components, transforming data via a modulus-like nonlinearity.

Conclusions:

• PNNs offer a viable platform for brain-inspired computation.
• The findings support the development of novel computational schemes leveraging PNN criticality.
• This research paves the way for efficient, low-power neuromorphic computing systems.