Comparing weak- and unsupervised methods for resonant anomaly detection

Abstract

Anomaly detection techniques are growing in importance at the Large Hadron Collider (LHC), motivated by the increasing need to search for new physics in a model-agnostic way. In this work, we provide a detailed comparative study between a well-studied unsupervised method called the autoencoder (AE) and a weakly-supervised approach based on the Classification Without Labels (CWoLa) technique. We examine the ability of the two methods to identify a new physics signal at different cross sections in a fully hadronic resonance search. By construction, the AE classification performance is independent of the amount of injected signal. In contrast, the CWoLa performance improves with increasing signal abundance. When integrating these approaches with a complete background estimate, we find that the two methods have complementary sensitivity. In particular, CWoLa is effective at finding diverse and moderately rare signals while the AE can provide sensitivity to very rare signals, but only with certain topologies. We therefore demonstrate that both techniques are complementary and can be used together for anomaly detection at the LHC.

Autoencoders for unsupervised anomaly detection in high energy physics

“…It is not surprising that the AE is not able to reconstruct the complex structure of top images when ignoring all but a few pixels. Limitations to the reconstruction of structures in high energy physics have also been noted recently in [51,52].…”

Machine Learning in the Search for New Fundamental Physics

“…In contrast to unsupervised searches, weakly-and semisupervised searches use some label information to inform the training. This can result in an improved sensitivity for BSM particles, at the cost of additional assumptions [139,140]. The aspect that distinguishes weakly supervised learning and semi-supervised learning is the fidelity of the labels.…”

Improving Variational Autoencoders for New Physics Detection at the LHC With Normalizing Flows

“…Recently, the use of machine learning techniques has been advocated as a mean to reduce the model dependence (Weisser and Williams, 2016 ; Collins et al, 2018 , 2019 , 2021 ; Blance et al, 2019 ; Cerri et al, 2019 ; D'Agnolo and Wulzer, 2019 ; De Simone and Jacques, 2019 ; Heimel et al, 2019 ; Andreassen et al, 2020 ; Cheng et al, 2020 ; Dillon et al, 2020 ; Farina et al, 2020 ; Hajer et al, 2020 ; Khosa and Sanz, 2020 ; Nachman, 2020 ; Nachman and Shih, 2020 ; Park et al, 2020 ; Amram and Suarez, 2021 ; Bortolato et al, 2021 ; D'Agnolo et al, 2021 ; Finke et al, 2021 ; Gonski et al, 2021 ; Hallin et al, 2021 ; Ostdiek, 2021 ). In this context, the particle-physics community engaged in two data challenges: the LHC Olympics 2020 (Kasieczka et al, 2021 ) and the DarkMachines challenge (Aarrestad et al, 2021 ), where different approaches were explored to attempt to detect an unknown signal of new physics hidden in simulated data.…”