| 1 |
%Due to the low signal rates for many new physics scenarios and large QCD
|
| 2 |
%backgrounds, efficiently identifying tau leptons while maintaining an extremely
|
| 3 |
%low mis-tag rate will be an important part of the CMS physics program. The CMS
|
| 4 |
%particle flow algorithm combines all sub-detectors to provide a global
|
| 5 |
%reconstruction of a collision event and potentially improves spatial and energy
|
| 6 |
%resolution. A new algorithm for identifying hadronic tau decays, the Tau Neural
|
| 7 |
%Classifier (TaNC) is presented in this paper. Using the reconstructed objects
|
| 8 |
%from particle flow, the algorithm reconstructs the hadronic decay of the tau
|
| 9 |
%lepton and uses an ensemble of neural nets to discriminate against common
|
| 10 |
%background. This strategy provides a large performance improvement with respect
|
| 11 |
%to previous CMS tau identification strategies and can potentially increase the
|
| 12 |
%reach of many CMS searches for physics beyond the Standard Model. A technical
|
| 13 |
%description of the algorithm and measurements of performance are included.
|
| 14 |
|
| 15 |
The Tau Neural Classifier (TaNC) is a novel algorithm for identification of
|
| 16 |
hadronic tau decays. The algorithm includes two components, the reconstruction
|
| 17 |
of tau lepton hadronic decay modes and discrimination of tau lepton hadronic
|
| 18 |
decays from quark and gluon jets. The reconstruction of decay modes is based
|
| 19 |
on the reconstruction of individual charged hadrons and photons by the
|
| 20 |
particle--flow algorithm and is utilized in the discrimination to train a set
|
| 21 |
of neural networks using input variables that are sensitive to particular decay
|
| 22 |
modes. We observe a significant improvement in identification performance in
|
| 23 |
comparison to previous algorithms.
|
