Kim A. Nicoli: "Deep Generative Models for Thermodynamics of Spin Systems and Field Theories"

Monday 16 Jan 2023, 16:15 → 17:15 Europe/Berlin

Nußallee 14-16/3.013 (HISKP) - Seminar Room II (HISKP)

Speaker: Kim A. Nicoli (TU Berlin)

Title: Deep Generative Models for Thermodynamics of Spin Systems and Field Theories

Abstract: In recent decades, machine learning has become a widely used tool in many fields of science, including Quantum Chemistry, Geophysics, Material Science, and Biophysics, among others. Revolutionary results, such as the predictive accuracy of Alphaphold 2, have shown that machine learning will play an increasingly important role in scientific research.

In this talk, I will present a new paradigm in which lattice field theory and machine learning are combined to achieve improved results. Specifically, I will introduce a general framework that enhances standard sampling algorithms by incorporating generative models, a subclass of machine learning techniques. I will demonstrate the application of this framework to two physical systems: a two-dimensional Ising model and a phi-4 lattice field theory. During the discussion, both the advantages and limitations of this approach will be identified and addressed, with the aim of developing a more solid and general framework.

Talk Bonn.pdf