Wing Sheung Chan

Object reconstruction and identification 57 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 RNN score 5 − 10 4 − 10 3 − 10 2 − 10 1 − 10 1 candidates / 0.02 had–vis τ Fraction of ATLAS Simulation Preliminary had–vis τ 1–prong had–vis τ True had–vis τ Fake (a) 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 RNN score 5 − 10 4 − 10 3 − 10 2 − 10 1 − 10 1 candidates / 0.02 had–vis τ Fraction of ATLAS Simulation Preliminary had–vis τ 3–prong had–vis τ True had–vis τ Fake (b) 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 efficiency had–vis τ True 1 10 2 10 3 10 4 10 rejection had–vis τ Fake ATLAS Simulation Preliminary RNN (1–prong) BDT (1–prong) Working points (1–prong) RNN (3–prong) BDT (3–prong) Working points (3–prong) (c) Figure 3.4.: The expected output distributions for (a) 1-prong and (b) 3-prong τ had - vis candi- dates, and (c) ROC curves of the RNN ID [82] . The markers on the ROC curves indicate the four defined working points, Tight , Medium , Loose and VeryLoose , with increasing signal selection efficiencies.