Wing Sheung Chan

Event selection and classification 75 Table 4.4.: Event selection for CRZ ττ , CRZ `` and the FRs, listing only the criteria that are different or additional to those for the SR. The selection criteria for CRZ `` involve outputs of neural network classifiers, which will be described in Section 4.2. Region Selection criteria eτ channel µτ channel CRZ τ τ m T ( τ had - vis , E miss T ) > 35 GeV m T ( e, E miss T ) < 40 GeV m T ( µ, E miss T ) < 40 GeV 70 GeV < m coll ( e, τ ) < 110 GeV 70 GeV < m coll ( µ, τ ) < 110 GeV CRZ `` NN Z `` output < 0 . 2 NN Z ττ output > 0 . 8 NN Wjets output > 0 . 8 FRW m T ( τ had - vis , E miss T ) > 35 GeV m T ( e, E miss T ) > 40 GeV m T ( µ, E miss T ) > 40 GeV FRQ Non-isolated light lepton candidate m T ( e, E miss T ) < 40 GeV m T ( µ, E miss T ) < 40 GeV SS charged e – τ had - vis pair SS charged µ – τ had - vis pair FRZ Two electrons and no muons Two muons and no electrons 81 GeV < m ( e 1 , e 2 ) < 101 GeV 81 GeV < m ( µ 1 , µ 2 ) < 101 GeV FRT Two or more b -jets 4.2.1. Training samples The NNs are trained using MC samples (see Section 5.1) selected by the same event selection criteria for the SR, except that the cuts on m vis ( `, τ ) and the NN outputs are omitted. In addition, real τ had - vis candidates from Z → `τ and Z → τ τ events are only required to pass the Loose instead of the Tight identification in order to increase the size of the training samples. For the Z → `` sample, events are selected only if the true origin of the τ had - vis candidate is a misidentified light lepton. Different NNs are separately trained for the eτ and µτ channels and for the 1P and 3P regions. To increase the signal sample size, the Z → eτ and Z → µτ samples are combined and used for training in both channels, assuming equivalent event topology when exchanging e and µ . Due to the small contribution of Z → `` events to the total background in the 3P regions, no classifier is trained to discriminate against them. The sizes of the training samples used for the NN training are shown in Table 4.5.

RkJQdWJsaXNoZXIy ODAyMDc0