When we physicists try to navigate the space of new particle physics theories we need to utilise all the information we can from a wide range of different experiments. This is the task of global fits, large-scale analyses that confront new theories with all relevant experimental data. A key limiting factor for such analyses is the computational expense — exploring a theory with only a handful of free parameters can require tens of millions of CPU hours, and even then we are forced to make several approximations in our statistical analysis. In the PLUMBIN’ project, which brings together researchers from statistics and theoretical physics, we are working on a multi-pronged approach to tackle these computational and statistical challenges. This includes work on emulators capable of on-the-fly training and prediction, hypothesis testing when Wilks’ theorem does not apply, and development of sampling algorithms tailored for profile likelihood analyses. In this talk I will discuss what particle physics global fits are, what computational and statistical challenges we face, and our ongoing research in the PLUMBIN' project.
Anders Kvellestad: Particle physics global fits: how not to get lost in theory space
Anders Kvellestad is a Researcher in theoretical physics at the Department of Physics and head of the GAMBIT Collaboration. His research focuses on how to test new theories against results from the Large Hadron Collider and other experiments, using both frequentist and Bayesian approaches, and the computational challenges related to this work. Anders completed his PhD at the University of Oslo in 2015 and has since held postdoctoral positions at Nordita (Stockholm), Imperial College London and UiO.
Since 2021 he has lead the GAMBIT Collaboration, an international collaboration of physicists working on global fit analyses. He is also part of the PLUMBIN’ project at UiO, which aims to develop new computational and statistical tools for particle physics and beyond.
