A Data-Directed Paradigm for BSM searches: the bump-hunting example

TL;DR

This paper introduces a data-directed paradigm (DDP) that leverages neural networks to efficiently identify deviations from the Standard Model in particle physics data without relying on simulations, enhancing bump-hunting methods.

Contribution

The paper presents a novel DDP that combines neural networks with bump-hunting, reducing reliance on simulations and enabling rapid, sensitive searches for new physics signals.

Findings

Neural networks can replace traditional background estimation methods.

The DDP maintains high sensitivity to bumps with minimal degradation.

The approach allows efficient testing of multiple final states.

Abstract

We propose a data-directed paradigm (DDP) to search for new physics. Focusing on the data without using simulations, exclusive selections which exhibit significant deviations from known properties of the standard model can be identified efficiently and marked for further study. Different properties can be exploited with the DDP. Here, the paradigm is demonstrated by combining the promising potential of neural networks (NN) with the common bump-hunting approach. Using the NN, the resource-consuming tasks of background and systematic uncertainty estimation are avoided, allowing rapid testing of many final states with only a minor degradation in the sensitivity to bumps relative to standard analysis methods.