Simulating the Quantum World with Data-Free, Physics-Driven Machine Learning

The behavior of electrons is chiefly responsible for the properties of materials and molecules. Predicting the behavior of many interacting electrons poses a significant scientific challenge and has led to the development of many methods of tackling problems in quantum many-body physics.

In this lecture, Giuseppe Carleo will focus on simulation-driven machine learning techniques. He’ll explore how artificial neural networks can represent quantum states, offering a powerful alternative to traditional variational methods. The talk will introduce how these approaches systematically and controllably learn many-body wave functions without relying on pre-existing data. He’ll examine applications in diverse domains, including condensed matter, chemistry and nuclear physics. Special attention will be given to how neural network representations have advanced our ability to simulate prototypical many-body quantum systems, surpassing previous variational descriptions.

Carleo is a professor of physics at the Swiss Federal Institute of Technology Lausanne (EPFL), where he leads the Computational Quantum Science Laboratory. He earned his Ph.D. from the International School for Advanced Studies (SISSA) in Italy in 2011 and held postdoctoral positions at the Institut d’Optique in France and ETH Zurich in Switzerland. From 2018 to 2020, he was a research scientist at the Flatiron Institute’s Center for Computational Quantum Physics in New York City. Carleo’s expertise is in developing advanced numerical algorithms for studying strongly interacting quantum systems. He pioneered the application of machine learning techniques to quantum physics, introducing neural network representations of quantum states. Carleo is also known for developing the time-dependent variational Monte Carlo method and leading the open-source project NetKet.