Unraveling Complex Magnetism with Brain-Inspired Computing

Research question

We use brain-inspired artificial neural networks to explore the complex phase and parameter spaces of atomistic spin models, which realistically describe magnetic materials. Specifically, we use Bayesian Confidence Propagating Neural Networks (BCPNN), developed by the WASP partner, for unsupervised learning. The purpose is two-fold: to benchmark these cutting-edge AI methods on a physically relevant problem, and to search for novel magnetic phases that are promising for developing next generation computing paradigms (e.g. neuromorphic computing) with extremely high energy efficiency.

Sustainability aspects

Surging energy consumption from the use of information technology, with accompanying CO2 emissions, calls for more energy efficient information processing. This is more urgent than ever with AI permeating all areas of our society. Our project addresses this challenge by aiming at the design of novel magnetic phases for future high-energy efficiency computing.