Skip to content

implement filter logic and integrate ML decision in trackoverlay

Fang-Ying Tsai requested to merge fatsai/athena:filterML_dev into main

This MR introduces changes to enhance the track-overlay workflow with ML decisions. The key updates include:

  1. Parallel sequences: the main adjustments are in the RecoSteering.py file. We then needed to update tool names to align with changes in parallel sequences for track-overlay workflows based on ML scores to prevent crashes.
  2. New Flag, flags.Reco.EnableTrackOverlay: added a new flag that allows users to enable the track-overlay workflow with ML decision-making.
  3. New Flag, flags.TrackOverlay.MLThreshold: the recommended threshold for optimal performance is 0.74201. Users can change the threshold by setting it through this flag.

The ML model has been uploaded to cvmfs. See JIRA for more details. Explanations regarding filters and flags can be found in the presentation slides 13, 14, and 15.

Tagging @jchapman @pagacova @lmijovic @rwang @schaarsc @mbandier @tlari @dduda @sjiggins @wleight

Edited by Fang-Ying Tsai

Merge request reports