Skip to content

New ghostprob NN weights for PbPb run (old)

Ivan Cambon Bouzas requested to merge jcambonb-changes into master

NN weights for ghostprob are being added for PbPb24 reconstruction. The data used was expected 2024 peripheral PbPb minimum bias MC and the training samples were generated using as base the hlt2_ghostprob_training_data.py but using the current PbPb reconstruction with UT. For training, the hlt2_ghostprob_training.py has been used with the same features as pp for each type of track. A plot with the ghost rate vs scifi hits is shown for Best Long tracks. Both reconstructions are standalone_hlt2_light_reco_pr_kf reconstruction, where in the dark points the pp NN weights are used while in the blue the NN weights used are the ones for PbPb. To add the new weights, changes to Moore have been made. This merge request is linked to Moore!3646

ghostrate_vs_FTHits_BestLong.pdf

Merge request reports