Calorimeter Machine Learning

Thursday 31 Aug 2017, 19:00 → 21:20 Europe/Zurich

Summary:

1. Matt presented the BDT results of the gamma vs. pi0 classifier. ECALmomentY5 still contributed more than ECALmomentX5 (without adding subjettiness).

2. Doubling the training sample size for electron vs. charged pion classifier to 400K did not increase the accuracy. Maybe we need more training data?

3. A faster generator will be released.

4. We also debugged the code for subjettiness.

1 Maurizio

Speaker: Maurizio Pierini (CERN)

2 Ben

Speaker: Benjamin Henry Hooberman (Univ. Illinois at Urbana-Champaign (US))

3 Amir

Speaker: Amir Farbin (University of Texas at Arlington (US))

4 Jean-Roch

Speaker: Dr Jean-Roch Vlimant (California Institute of Technology (US))

5 Ryan

Speaker: Ryan Reece (University of California,Santa Cruz (US))

6 Matt

Speaker: Matt Zhang (Univ. Illinois at Urbana-Champaign (US))

7 Taylor