Taus decay via the weak interaction, necessitating the production of a neurino system. Leptonic tau decays occur 35% of the time, producing a single light lepton (an electron or muon) and a system of two neutrinos. Hadronic tau decays occur 65% of the time, producing a hadronic shower and a single neutrino. The neutrino systems in tau decays are invisible to detectors like CMS; accordingly, when a Higgs boson decays to two taus, the mass spectrum of the measured (visible) tau decay products typically undershoots the mass of the original ditau system. The fastmtt algorithm was developed to correct the mass spectrum of a visible ditau decay system to account for missing energy due to neutrino production.
This code implements the algorithm presented in http://cms.cern.ch/iCMS/jsp/openfile.jsp?tp=draft&files=AN2019_032_v3.pdf. A dedicated C++ implementation of this algorithm is available here: https://github.com/SVfit/ClassicSVfit/tree/fastMTT_19_02_2019.
The repo is currently organized as follows:
test_fastmtt.py*contains a Numba-accelerated Pythonic implementation of fastmtt; to test this implementation, run the following command:python test_fastmtt.py <input sample> <n_evts>fastmtt_classic/contains the original C++ implementation of fastmtt; to build a shared library importable in Python, runpython build_fastmtt_classic.pyfrom this directorysamples/contains a set of test samples skimmed and formatted for easy input to fastmttnotebooks/contains a set of notebooks focused on validating the performance of this fastmtt implementation against the classic C++ version