Séminaire IP2I: Automatic BSM model parameter space exploration with Active Learning

Séminaire réalisé par Mark Goodsell – LPTHE

For a theorist to confront a given model or scenario to the latest limits requires a whole chain of tools and calculations to map from Lagrangian parameters to data. With the immense range of possible models now under consideration this begs for automation and genericity. I will review the state of the art of this toolchain, upcoming developments and the current challenges within the context of the SARAH package, which allows phenomenologists to go from any renormalisable theory defined at any energy scale to precision computations and collider tools at the push of a button. However, running this chain can often be computationally expensive, and given that models often have a parameter space with a large dimensionality, it is necessary when trying to explore a new model to choose an efficient algorithm for selecting which parameter points to analyse. I will describe how Active Learning can be such an algorithm and show some applications such as finding the largest possible dark matter mass.

IP2I seminars are open to all UCBL and CNRS students and collaborators.
More information on seminars: https://intranet.ip2i.in2p3.fr/vie-scientifique/seminaires-ip2i