Göttingen 2025 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 11: Data, AI, Computing, Electronics I (Statistical Methods, Applications)
T 11.3: Vortrag
Montag, 31. März 2025, 17:15–17:30, VG 2.101
Making your analysis reusable with model-agnostic likelihoods and their serialization — •Lorenz Gärtner1, Thomas Kuhr1, Slavomira Stefkova2, Danny van Dyk3, Lukas Heinrich4, Méril Reboud5, Nikolai Krug1, and Malin Horstmann4 — 1LMU, Munich, Germany — 2University of Bonn, Bonn, Germany — 3IPPP Durham, United Kingdom — 4Technical University Munich, Germany — 5Université Paris-Saclay, France
What constitutes a "signal" in particle physics? Typically, a signal is defined by a specific physical process of interest. Using simulations, we approximate the probability densities of such processes and compare them to known backgrounds through likelihood-based methods.
What happens when parameters for the process need to be revised? What if more precise theoretical predictions become available? How can we search for an entire class of similar processes parameterized by multiple variables? Moreover, can we leverage results from multiple analyses to constrain these parameters?
We address these questions with a simple and efficient reinterpretation approach. We construct model-agnostic likelihoods by employing kinematic reweighting techniques, enabling flexible exchanges of signal models and inference on underlying physical parameters. This method’s generality ensures compatibility across analyses, while its straightforward serialization facilitates easy distribution and reuse.
To demonstrate the power and simplicity of our approach, we use the likelihoods of the Belle II B+ → K+ ν ν analysis to constrain theory parameters.
Keywords: reinterpretation; statistics; belle 2; likelihood; kinematics