Session L02: Generative Models for Fundamental Physics

Generative Models for Lattice Field Theory

In the context of lattice quantum field theory calculations in particle and nuclear physics, I will describe avenues to accelerate sampling from known probability distributions using machine learning. I will focus in particular on flow-based generative models, and describe how guarantees of exactness and the incorporation of complex symmetries into model architectures can be achieved. I will show the results of proof-of-principle studies that demonstrate that sampling from generative models can be orders of magnitude more efficient than traditional sampling approaches such as Hamiltonian/hybrid Monte Carlo in this context, and discuss the potential impacts of these approaches in nuclear and particle physics.   

Generative Models for Particle Physics

Over the next years, measurements at the LHC and the HL-LHC will provide us with a wealth of data. The best hope of answering fundamental questions like the nature of dark matter, is to adopt big data techniques for precision simulations and optimized analyses to extract all relevant information.

LHC physics relies at a fundamental level on our ability to simulate events efficiently from first principles. In the coming LHC runs, these simulations will face unprecedented precision requirements to match the experimental accuracy. Generative models have become a central tool to overcome limitations from high precision in event generation and high dimensionality of detector simulations. Such networks can be employed within established simulation tools, as part of a simulation frameworks, or to compress measured data. Recent studies have demonstrated that generative networks can achieve high-precision in simulations while maintaing control over training stability and associated uncertainties.

Since generative networks in the form of normalizing flows can be inverted, they also open new avenues in LHC analyses. The access to the density of the generated distribution enables new methods for anomaly detection, while their interpretation in terms of probability densities leads to new methods for multi-dimensional unfolding.