Karlsruhe 2024 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

T: Fachverband Teilchenphysik

T 2: Search for new particles 1 (LHC)

T 2.3: Talk

Monday, March 4, 2024, 16:30–16:45, Geb. 20.30: 1.067

Searching for new physics in dijets using anomaly detection — •Manuel Sommerhalder, Tobias Quadfasel, Louis Moureaux, and Gregor Kasieczka — Institut für Experimentalphysik, Universität Hamburg

Despite compelling experimental and theoretical motivation as well as extensive new physics searches at the Large Hadron Collider, there have been no discoveries of physics beyond the standard model (BSM) to date. A potential reason for this might be that the common search strategy relies on selecting BSM signal candidate events based on specific signal and background models. Such a dedicated search cannot be performed for every possible BSM theory and phase space region. And even if this was computationally feasible, it would still lack sensitivity to unthought-of models. Thus, model-independent anomaly detection methods are an important addition to the existing search paradigm. These algorithms aim to select signal candidates in a data-driven manner based on anomalous phase space signatures.

One such anomaly detection method is CATHODE. It detects resonant signal peaks by combining neural density estimation in a sideband region with a weakly supervised classification task of distinguishing real data from an in-situ simulation of the background. We present the first application of CATHODE in a search for BSM physics in the CMS experiment targeting a dijet final state.

Keywords: anomaly detection; dijet; CMS; model-agnostic; machine learning