From Machine Learning to Computational Quantum Physics

Abstract

Machine-learning techniques that originated in several domains of “Artificial Intelligence”, are increasingly being applied to tackle problems in physical sciences. Applications range from experimental fields, where ML has been leveraged to more efficiently carry out measurements, inhibit noise or design structures, to computational and theoretical physics. In particular, a recent breakthrough showed that several numerical simulation tasks, such as the search for the ground-state of an Hamiltonian or the simulation of the Schrödinger’s equation, can be recast as Machine-Learning problems and efficiently their solution can be efficiently encoded with Neural-Networks.

I will begin this talk with a brief overview of modern applications of ML to quantum sciences. Then, I will focus on computational physics, introducing Neural-Network Quantum States and discussing recent advancements in the simulation of strongly-correlated materials, possibly giving a practical “hands-on” examples.

References

• https://arxiv.org/abs/2204.04198 (Chapter 5)