Skip to main content

Table 3 95% CL upper limit on the signal strength \(\mu \) of each benchmark signal for the different AD methods using the full feature set and the sanitised set and for a dedicated supervised DNN model trained on the full feature set. The statistical uncertainties, including the effect from limited statistics in the simulated datasets, are also shown

From: Finding new physics without learning about it: anomaly detection as a tool for searches at colliders

Model Benchmark signal
  FCNC HG No HG
   1.0 TeV 1.2 TeV 1.4 TeV 1.0 TeV 1.2 TeV 1.4 TeV
Full features
Supervised DNN \(7^{+3}_{-2}\) \(0.05^{+0.06}_{-0.03}\) \(0.019^{+0.010}_{-0.007}\) \(0.013^{+0.008}_{-0.004}\) \(0.03^{+0.02}_{-0.01}\) \(0.12^{+0.06}_{-0.04}\) \(0.4^{+0.2}_{-0.2}\)
\(H_T\) \(100^{+60}_{-20}\) \(0.14^{+0.07}_{-0.05}\) \(0.16^{+0.08}_{-0.06}\) \(0.16^{+0.08}_{-0.05}\) \(0.4^{+0.3}_{-0.1}\) \(1.0^{+0.5}_{-0.3}\) \(1.8^{+0.9}_{-0.6}\)
Deep SVDD \(10^{+8}_{-1}\) \(0.15^{+0.07}_{-0.04}\) \(0.17^{+0.08}_{-0.05}\) \(0.21^{+0.09}_{-0.07}\) \(0.3^{+0.2}_{-0.1}\) \(1.1^{+0.5}_{-0.3}\) \(3.1^{+1.4}_{-0.9}\)
AE \(30^{+20}_{-8}\) \(0.029^{+0.014}_{-0.009}\) \(0.03^{+0.02}_{-0.01}\) \(0.04^{+0.02}_{-0.01}\) \(0.06^{+0.03}_{-0.02}\) \(0.21^{+0.10}_{-0.07}\) \(0.6^{+0.3}_{-0.2}\)
HBOS \(100^{+60}_{-20}\) \(0.15^{+0.07}_{-0.05}\) \(0.17^{+0.08}_{-0.05}\) \(0.19^{+0.09}_{-0.06}\) \(0.4^{+0.1}_{-0.1}\) \(1.0^{+0.5}_{-0.3}\) \(2.7^{+1.3}_{-0.9}\)
iForest \(100^{+100}_{-3}\) \(0.19^{+0.10}_{-0.06}\) \(0.23^{+0.12}_{-0.08}\) \(0.26^{+0.14}_{-0.09}\) \(0.5^{+0.2}_{-0.2}\) \(1.4^{+0.7}_{-0.5}\) \(4^{+2}_{-2}\)
Sanitised features
Supervised DNN \(6^{+3}_{-2}\) \(0.008^{+0.004}_{-0.003}\) \(0.009^{+0.005}_{-0.003}\) \(0.006^{+0.003}_{-0.001}\) \(0.009^{+0.005}_{-0.003}\) \(0.04^{+0.03}_{-0.01}\) \(0.3^{+0.2}_{-0.1}\)
\(H_T\) \(100^{+50}_{-30}\) \(0.14^{+0.07}_{-0.05}\) \(0.16^{+0.08}_{-0.06}\) \(0.16^{+0.08}_{-0.05}\) \(0.4^{+0.3}_{-0.1}\) \(1.0^{+0.5}_{-0.3}\) \(1.8^{+0.9}_{-0.6}\)
Deep SVDD \(10^{+7}_{-2}\) \(0.08^{+0.04}_{-0.02}\) \(0.08^{+0.04}_{-0.02}\) \(0.09^{+0.04}_{-0.02}\) \(0.15^{+0.06}_{-0.04}\) \(0.5^{+0.2}_{-0.1}\) \(1.4^{+0.6}_{-0.4}\)
AE \(100^{+100}_{-1}\) \(0.0053^{+0.0006}_{-0.0005}\) \(0.0068^{+0.0006}_{-0.0006}\) \(0.0089^{+0.0007}_{-0.0008}\) \(0.0104^{+0.0009}_{-0.0012}\) \(0.042^{+0.004}_{-0.004}\) \(0.15^{+0.02}_{-0.01}\)
HBOS \(100^{+60}_{-20}\) \(0.19^{+0.11}_{-0.06}\) \(0.21^{+0.13}_{-0.07}\) \(0.22^{+0.14}_{-0.07}\) \(0.4^{+0.2}_{-0.1}\) \(1.1^{+0.6}_{-0.4}\) \(2.7^{+1.7}_{-0.9}\)
iForest \(100^{+100}_{-5}\) \(0.18^{+0.09}_{-0.06}\) \(0.19^{+0.09}_{-0.07}\) \(0.19^{+0.09}_{-0.07}\) \(0.4^{+0.3}_{-0.1}\) \(1.0^{+0.6}_{-0.3}\) \(2.4^{+1.2}_{-0.8}\)