Contributor info: Prof. Rebeca Gonzalez Suarez

ATLAS Collaboration

SciPost Phys. 19, 155 (2025) · published 16 December 2025 | Toggle abstract · pdf

The ATLAS experiment at the Large Hadron Collider explores the use of modern neural networks for a multi-dimensional calibration of its calorimeter signal defined by clusters of topologically connected cells (topo-clusters). The Bayesian neural network (BNN) approach not only yields a continuous and smooth calibration function that improves performance relative to the standard calibration but also provides uncertainties on the calibrated energies for each topo-cluster. The results obtained by using a trained BNN are compared to the standard local hadronic calibration and to a calibration provided by training a deep neural network. The uncertainties predicted by the BNN are interpreted in the context of a fractional contribution to the systematic uncertainties of the trained calibration. They are also compared to uncertainty predictions obtained from an alternative estimator employing repulsive ensembles.

Olga Sunneborn Gudnadottir, Axel Gallén, Giulia Ripellino, Jochen J. Heinrich, Raazesh Sainudiin, Rebeca Gonzalez Suarez

SciPost Phys. 18, 080 (2025) · published 4 March 2025 | Toggle abstract · pdf

This study presents a novel method for the definition of signal regions in searches for new physics at collider experiments. By leveraging multi-dimensional histograms with precise arithmetic and utilizing the SparkDensityTree library, it is possible to identify high-density regions within the available phase space, potentially improving sensitivity to very small signals. Inspired by a search for dark mesons at the ATLAS experiment, CMS open data is used for this proof-of-concept intentionally targeting an already excluded signal. Signal regions are defined based on density estimates of signal and background. These preliminary regions align well with the physical properties of the signal while effectively rejecting background events.