Purpose: Test potential of ML using multiple subdetectors, especially in the case of visible signal.
- 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 hassympy
for A'->e+e- production), or make your own.
- You can pull
- If using v2.X.X, change
libFramework.so
tolibEvent.so
inmainframe.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 inROOT.gSystem.Load()
inmainframe.py
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
andmods/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>
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.)