Machine-Learning-Based Sampling in Lattice Field Theory and Quantum Chemistry
October 21-25, 2024 (Bethe Forum)
Bethe Center for Theoretical Physics, Bonn, Germany
The application of generative machine-learning models for sampling Boltzmann distributions has first been proposed in the context of quantum chemistry, and has later been successfully applied in lattice quantum field theory. Both research fields have quickly and independently progressed during the last years. With this workshop, we aim to bring together researchers from both communities, in order to foster exchange on common challenges and opportunities, including
- inductive biases and symmetries: integration of physical symmetries and prior knowledge into generative models for better efficiency,
- mode coverage: efficient sampling from multimodal distributions,
- scalability: scaling up to larger physical and chemical system sizes,
- interpretability: analyzing the capability of a trained model to sample from a target distribution and to correctly capture the system properties,
- uncertainty estimation: reliable estimation of both statistical and systematic uncertainties,
- software: development of open-source code and software packages for machine-learning-based sampling.
Workshop Structure
The workshop program includes
- 3 keynote talks,
- 3 panel discussions,
- 14 technical talks,
- 1 poster session,
- plenty of room for discussions.
Two colloquia on related topics are also planned during the week of the workshop.
We strongly encourage early-career researchers, including students and early postdocs, to present their research at the poster session. The posters will be advertised during a lightning talks session. The best poster will be awarded a poster prize.
Organizers
- Kim A. Nicoli (University of Bonn)
- Lena Funcke (University of Bonn)
- Tom Froembgen (University of Bonn)
- Shinichi Nakajima (Technical University of Berlin)
This workshop is funded by the Bethe Center for Theoretical Physics (BCTP) and the Transdisciplinary Research Area "Building Blocks of Matter and Fundamental Interactions" (TRA Matter).
