Fig. 3: Physics performance results.
From: Nanosecond anomaly detection with decision trees and real-time application to exotic Higgs decays
The distribution is given for anomaly scores Δ (left column) and the ROC curves (right column) for the \(H\to a{a}^{{\prime} }\to \gamma \gamma jj\) scenario (top row) and the LHC physics dataset83 (bottom row). Along with the ROC curves for the γγjj dataset (top right), the operating points of the \({p}_{{{{{{{{\rm{T}}}}}}}}}^{\gamma 2} \, > \) 25 GeV trigger are shown, with numerical values to compare it to the autoencoder’s performance. Values shown are fractions of all events in the sample. The autoencoder is trained only on the respective Standard Model (SMγγjj and SMcocktail) processes. TPR and FPR represent true and false positive rates, respectively. The plots are software-simulated results using bit integers as done in the firmware.
