SMuK 2023 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 131: Searches V

T 131.5: Vortrag

Donnerstag, 23. März 2023, 18:30–18:45, HSZ/0403

QCD Generative Model Without Machine Learning — •Samuel Bein — Universität Hamburg, Hamburg, Germany

The Rebalance and Smear technique for the modeling of QCD backgrounds to searches for dark matter at the LHC is presented as a publicly available toolkit. Bayesian inference is carried out on real data events to estimate a latent space of the true jet energy values within each event. The latent space is sampled multiple times per event according to a known PDF of the detector response to the jet energy, and the resulting collection represents a high-statistics proxy for the true QCD background. This method, previously carried out at CMS and ATLAS for background estimation, can be further employed in the training of multivariate classifiers to optimally extend the sensitivity of searches to BSM scenarios with compressed mass spectra. An example future search probing pure Higgsino dark matter in gluino and squark simplified models, is a suitable application of this method in Run 3.