Autoencoders for Real-Time SUEP Detection

TL;DR

The paper tackles real-time detection of Soft Unclustered Energy Patterns (SUEP) from hidden-valley dark sectors at the LHC by reframing it as an anomaly-detection problem in the CMS High-Level Trigger (HLT).It develops a deep convolutional autoencoder (ConvAE) that operates on three-channel ET maps from the inner tracker, ECAL, and HCAL, trained in an unsupervised fashion on QCD backgrounds and using a sparsity-aware loss based on the inverse Dice Loss.Key results show that the ConvAE can identify a subset of SUEP events with around 40% efficiency while maintaining a low QCD mistag rate (as low as 2% for some mass hypotheses) and achieving CPU inference times around 20 ms, compatible with HLT latency constraints.Because the method is unsupervised, it is model-agnostic and readily adaptable to other new-physics signatures that produce atypical detector signatures, with potential applicability beyond CMS.

Abstract

Confining dark sectors with pseudo-conformal dynamics can produce Soft Unclustered Energy Patterns (SUEP), at the Large Hadron Collider: the production of dark quarks in proton-proton collisions leading to a dark shower and the high-multiplicity production of dark hadrons. The final experimental signature is spherically-symmetric energy deposits by an anomalously large number of soft Standard Model particles with a transverse energy of O(100) MeV. Assuming Yukawa-like couplings of the scalar portal state, the dominant production mode is gluon fusion, and the dominant background comes from multi-jet QCD events. We have developed a deep learning-based Anomaly Detection technique to reject QCD jets and identify any anomalous signature, including SUEP, in real-time in the High-Level Trigger system of the Compact Muon Solenoid experiment at the Large Hadron Collider. A deep convolutional neural autoencoder network has been trained using QCD events by taking transverse energy deposits in the inner tracker, electromagnetic calorimeter, and hadron calorimeter sub-detectors as 3-channel image data. Due to the sparse nature of the data, only ~0.5% of the total ~300 k image pixels have non-zero values. To tackle this challenge, a non-standard loss function, the inverse of the so-called Dice Loss, is exploited. The trained autoencoder with learned spatial features of QCD jets can detect 40% of the SUEP events, with a QCD event mistagging rate as low as 2%. The model inference time has been measured using the Intel CoreTM i5-9600KF processor and found to be ~20 ms, which perfectly satisfies the High-Level Trigger system's latency of O(100) ms. Given the virtue of the unsupervised learning of the autoencoders, the trained model can be applied to any new physics model that predicts an experimental signature anomalous to QCD jets.

Figures (9)

• Figure 1: $E_{\mathrm{T}}$ deposits for a QCD event in the inner tracker (left), ECAL (middle), and HCAL (right).
• Figure 2: $E_{\mathrm{T}}$ deposits for a SUEP(1000$\,\text{Ge\spaceV}$) event in the inner tracker (left), ECAL (middle), and HCAL (right).
• Figure 3: Comparisons of true $E_{\mathrm{T}}$ (left) and reconstructed $E_{\mathrm{T}}$ (right) (sum over all the pixels) between QCD and SUEP(1000 $\,\text{Ge\spaceV}$).
• Figure 4: Combined true $E_{\mathrm{T}}$ (left) and reconstructed $E_{\mathrm{T}}$ (right) for a QCD event in the inner tracker, ECAL, and HCAL.
• Figure 5: Combined true $E_{\mathrm{T}}$ (left) and reconstructed $E_{\mathrm{T}}$ (right) for a SUEP(1000 $\,\text{Ge\spaceV}$) event in the inner tracker, ECAL, and HCAL.