Göttingen 2025 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
T: Fachverband Teilchenphysik
T 46: Higgs Physics VI (top-Higgs Coupling)
T 46.8: Talk
Wednesday, April 2, 2025, 18:00–18:15, ZHG105
Application of the JAX-based Statistical Tool Evermore to a CMS Higgs Analysis — Peter Fackeldey2, •Felix Zinn1, Benjamin Fischer1, and Martin Erdmann1 — 1RWTH Aachen University — 2Princeton University
For precise measurements of the Higgs boson cross sections and coupling strengths, a likelihood based approach is typically needed for statistical inference. We introduce the python software package evermore which allows to define corresponding likelihood functions.
It is purely based on JAX and thus enables novel computing concepts such as automatic differentation and vectorization in the context of likelihood fitting.
We show how to build and evaluate a likelihood function in evermore with the example of an analysis of the tH and ttH production channel.
Furthermore we present how to set an upper limit on a parameter of interest. This procedure often requires the generation of pseudo data. We show how vectorized computation in evermore can be used for a toy-based approach.
Keywords: evermore; statistical inference; likelihoods; limits