Generative and variational modeling for quantum many-body physics

Giuseppe Carleo
Flatiron Institute, a Division of the Simons Foundation
Center for Computational Quantum Physics

I will present applications of machine learning techniques to the realm of many-body quantum physics, discussing challenges and successes obtained in the past few years.
First, I will discuss the central object to be modeled, the many-body wave function, and its parameterization in terms of artificial neural networks [1].
I will then introduced the concept of variational learning, naturally emerging in quantum physics, and in a middle ground between generative modeling and reinforcement learning.
In this context, I will present recent extensions of autoregressive generative models, particularly suitable for variational learning and other tasks [2].
I will finally discuss several applications suitable for experimental data, including Quantum State Tomography of highly-entangled states [3].

References:
[1] Carleo, and Troyer — Science 355, 602 (2017).
[2] Sharir, Levine, Wies, Carleo, and Shashua — arXiv:1902.04057
[3] Torlai, Mazzola, Carrasquilla, Troyer, Melko, and Carleo — Nature Physics 14, 447 (2018).


Back to Workshop I: From Passive to Active: Generative and Reinforcement Learning with Physics