Speaker
Description
CURTAINs is a fully data driven technique for creating background templates for use in searches for new physics processes. We employ invertible neural networks to learn the transformation to map any data point from its value of the resonant variable (e.g., invariant mass) to another chosen value. This conditional transformation allows us to create a template in the signal window, by mapping the data from the sidebands into the signal region. We demonstrate the effectiveness of this method by applying it to Anomaly Detection in a collider physics experiment and use it to enhance the sensitivity to new physics in a bump hunt. Using the LHCO dataset, we demonstrate that CURTAINs is competitive with other leading approaches, which aim to improve the sensitivity of bump hunts and can be trained on a much smaller range of the invariant mass. The method is applicable to any search for a resonant signal, including those resulting from dark interactions.
