Oscillating Neural Networks

Performing pattern recognition and solving complex optimization problems with coupled oscillator networks.

Overview

Combinatorial optimization problems find applicability in multiple industrial applications, such as drug synthesis, resource allocation, AI, computer vision, and circuit design. Solving these problems poses a persistent challenge for computer scientists today, as even the most advanced chips performance falls short compared to the capabilities of the human brain. We explore novel computing approaches emulating the brain's operations to solve optimization problems with high-power efficiency. We design oscillating neural networks driven by nanoscale switches harnessing vanadium dioxide's metal-insulator transition. Through rigorous device testing and circuit simulations, we power applications in neuromorphic computing and machine learning, bringing a touch of nature-inspired innovation to modern technology.

Neural Network based on VO2 oscillator Neurons and Analog Memristor synapses.

Current EU Project

Phastrac

Previous EU Projects

Phase-Change Switch
NeurONN
MANIC

SEM image of a VO2 crossbar device.

Publications

Growing smooth polycrystalline VO₂ film on silicon platform for neuromorphic computing
- - Olivier Maher
  - Roy Bernini
  - et al.
- 2022
- MRS Fall Meeting 2022
Coupled VO₂ Oscillators Circuit as Analog First Layer Filter in Convolutional Neural Networks
- - Elisabetta Corti
  - Joaquin Antonio Cornejo Jimenez
  - et al.
- 2021
- Frontiers in Neuroscience

Contributors

Siegfried Karg

Valeria Bragaglia

Olivier Maher