Resonant Anomaly Detection with Multiple Reference Datasets

TL;DR

This paper extends resonant anomaly detection techniques to utilize multiple reference datasets, enhancing detection performance and providing finite-sample guarantees, thus advancing high energy physics data analysis methods.

Contribution

It introduces generalized methods based on weak supervision that leverage multiple reference datasets, improving anomaly detection and offering finite-sample guarantees.

Findings

Improved detection performance with multiple datasets

Enhanced theoretical guarantees for finite samples

Validated on realistic and synthetic data

Abstract

An important class of techniques for resonant anomaly detection in high energy physics builds models that can distinguish between reference and target datasets, where only the latter has appreciable signal. Such techniques, including Classification Without Labels (CWoLa) and Simulation Assisted Likelihood-free Anomaly Detection (SALAD) rely on a single reference dataset. They cannot take advantage of commonly-available multiple datasets and thus cannot fully exploit available information. In this work, we propose generalizations of CWoLa and SALAD for settings where multiple reference datasets are available, building on weak supervision techniques. We demonstrate improved performance in a number of settings with realistic and synthetic data. As an added benefit, our generalizations enable us to provide finite-sample guarantees, improving on existing asymptotic analyses.