Python Implimentation of BDT

Purpose: Test potential of ML using multiple subdetectors, especially in the case of visible signal.

Requirments: Working install of ldmx-sw-v2.3.0 or greater.

• Commented version of rsegcont feature set requires the awkward python package which is not the standard ldmx container
  • You can pull jmlazaro/pro:vissig from docker hub (also has sympy for A'->e+e- production), or make your own.
• If using v2.X.X, change libFramework.so to libEvent.so in mainframe.py.
• confs/gabreille_back_v1.ini is used here as an example config. See its comments for further explanation.

In the directory containing ldmx-sw, $LDMX_BASE, enter source ldmx-sw/scripts/ldmx-env.sh.

• You may clone ldmx-sw with a different name (e.g. ldmx-sw-v3.0.0), just change it in the above command and in ROOT.gSystem.Load() in mainframe.py

Interactive

To make flat trees from ldmx event files:

ldmx python3 mainframe.py trees confs/<test.ini>

• -m can be used to give maximum number of events to run over
• -p can be used to selected a subset of processeses listed in the config file
• Other options can be found in mods/ROOTmanager.py and mods/configuration.py

To train BDT: ldmx hadd flat trees into the paths given in the config file

ldmx python3 mainfraim.py train confs/<test.ini>

To evaluate trained BDT on test samples:

ldmx python3 mainframe.py eval confs/<test.ini>

Batch

Remove ldmx and replace maineframe.py in corresponding interactive commands with batch.py.
(Some options are removed at this stage to avoid making big mistakes; add them at your own risk.)