What is it about?
Showing that a set of coupled oscillators can be trained to perform (as a first example) AI classification tasks (distinguishing images of different digits) if their couplings are trained using the 'Equilibrium Propagation' algorithm. We show that this works, at least in simulations, and is robust to various sources of experimental noise and finite precision. The next step is to try and build it in hardware!
Featured Image
Photo by Rachel Loughman on Unsplash
Why is it important?
The new types of computer such as these that exploit physical dynamics to do their computation can be massively more fast and energy-efficient that the GPUs currently used to train AI models - this is a big issue as huge sums of energy are currently being used to train fronteir models.
Read the Original
This page is a summary of: How to Train an Oscillator Ising Machine using Equilibrium Propagation, July 2025, Institute of Electrical & Electronics Engineers (IEEE),
DOI: 10.1109/icons69015.2025.00042.
You can read the full text:
Contributors
The following have contributed to this page
