Machine Learning LHC likelihoods

About

Full statistical models encapsulate the complete information of an experimental result, including the likelihood function given observed data. Their proper publication is of vital importance for a long lasting legacy of the LHC. Major steps have been taken towards this goal; a notable example being ATLAS release of statistical models with the pyhf framework. However, even the likelihoods are often high-dimensional complex functions that are not straightforward to parametrize. Thus, we propose to describe them with Normalizing Flows, a modern type of generative networks that explicitly learn the probability density distribution. As a proof of concept we focused on two likelihoods from global fits to SM observables and a likelihood of a NP-like search, obtaining great results for all of them. Complementarily, for New Physics search reinterpretation we are often only interested in the profiled likelihood given a signal strength, reducing the problem to a much less dimensional one. In this talk, we also discuss ongoing efforts on parametrising profiled likelihoods with neural networks..

0:00:00 Slide 1

0:03:46 Slide 2

0:04:57 Slide 3

0:08:24 Slide 4

0:09:42 Slide 5

0:11:04 Slide 6

0:13:14 Slide 7

0:15:05 Slide 8

0:18:38 Slide 9