Simulation-Based Inference with Uncertainty Quantification using Generative Models in Quantum Chromodynamics

Katherine Keegan, Emil Vatai, Aleksandr Drozd, Pi-Yueh Chuang, Xingfu Wu, nobuo sato, Johann Rudi, Mohamed Wahib

Keywords: Non-differentiable Parameter Inference, GANs, Quantum Correlation Functions

TL;DR: Parameter inference with GANs:enables uncertainty quantification, and is free of assumptions on parameters.

Abstract: Generative and adversarial machine learning methods have been used for parameter inference of physical models from observed data in various works. However, many real-world problems of interest involve non-differentiable models, a context in which many approaches cease to be sufficient. An example of this can be found in quantum chromodynamics, where inferring quantum correlation functions from observed data is hindered by the problem's intrinsic non-differentiability and stochasticity. To overcome this, we present a framework based fundamentally on generative adversarial networks in which parameters are iteratively optimized to generate realistic samples. This framework is novel compared to related works in that it simultaneously circumvents non-differentiability, enables uncertainty quantification, and is free of assumptions on parameters. We demonstrate the utility of this framework in learning synthetic distributions and simulated quantum correlation functions.

Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 8342