Machine Learning for Quantum Many-Body Physics
CECAM-FR-GSO, Le Village by CA Toulouse Évenement, Toulouse
Registration is still open.
The school will take place in a hybrid format (onsite and online).
Besides registration on this CECAM website, we ask you to also please register at https://mlqmb.sciencesconf.org (and follow the procedure there).
We aim at proposing a series of lectures and tutorials on applications of machine learning techniques to quantum physics, and in particular to quantum many-body physics.
Topics include introduction to machine learning in many-body physics, supervised and unsupervised learning, various forms of neural networks and their applications in quantum state reconstruction, Variational Monte Carlo etc.
Beyond traditional applications in pattern recognition or suggestion algorithms clearly changing our everyday life, the more recent sophisticated methods in machine learning have beaten problems judged so far as extremely complex (eg. AlphaGo for go playing). It is then with no surprise that all fields of science have been infused with AI and machine learning techniques.
Recently, researchers interested in quantum matter and the quantum many-body problem, which is often considered as a hard cookie due to its exponential complexity, have turned their attention to the algorithms underlying modern machine learning with remarkable success stories: automatic recognition of phases of matter , new generic variational ansatz competitive with the best wave-functions known so far , suggestion of new Monte Carlo moves leading to very efficient sampling , quantum state tomography , new ideas to fight the sign problem in quantum Monte Carlo  The field at the intersection between machine learning and hard quantum matter is clearly exploding and some authors think it’s a real revolution in numerical quantum physics [6,7]. Recent reviews highlight the strong impact of AI techniques in physics in general  and condensed matter in particular .
Our goal is to initiate a large audience of young European researchers to this new type of approach, which is very well developed in the US/Canada but still only sparsely represented in Europe. There were a few isolated initial school/conferences on this topic in Europe in 2019 and early 2020, but we feel it is the good time to set an entry point in the field for a larger number of researchers.
Lectures given by experts on this new area will introduce the main concepts and methods, and tutorials will be the opportunity to put this knowledge into practice. The lecturers have been chosen from the most active groups in the field. A special tutorial session will be devoted to learning the open source code NetKet.
UPDATE (04/02/2022) :
The school takes place in a hybrid (online/onsite) format.
Giuseppe Carleo (EPF Lausanne)
Mohamed Hibat Allah (Waterloo)
Florian Marquardt (Max Planck & University of Erlangen-Nuremberg)
Stefanie Czischek (Waterloo)
Evert van Nieuwenburg (Niels Bohr Institute, Copenhagen)
Filippo Vicentini (EPF Lausanne)
More to come...
Fabien Alet (CNRS) - Organiser
Sylvain Capponi (Univ. Paul Sabatier, Toulouse) - Organiser