Merging HLT1 NN ghost killer with forward_then_matching_no_ut optimisation

Alessandro Scarabottorequested to merge
ascarabo_merge_ghost_opt into master

Built on top of !1385 (closed) adding the NN ghost killer from !1405 (closed).

Plots using /scratch/allen_data/mdf_input/2024-expected-mdfs/mdf_sim-20231017-vc-md100/BsPhiPhi.mdf using the latest master which includes the improvement on 2024 MC from !1408 (merged).

I think it's the first time I see ~95% efficiency on long tracks with 4-5% ghost!

@dovombru @tevans @jzhuo

Forward no UT: efficiency Forward no UT: ghost
ForwardEffptLongfromB_2_eta_5_noUT_DEV_forward_withNN.pdf ForwardGhostRate_pt_noUT_DEV_forward_withNN.pdf
ForwardEffpLongfromB_2_eta_5_noUT_DEV_forward_withNN.pdf ForwardGhostRate_p_noUT_DEV_forward_withNN.pdf
Seeding&Matching: efficiency Seeding&Matching: ghost
ForwardEffptLongfromB_2_eta_5_noUT_DEV_matching_withNN.pdf ForwardGhostRate_pt_noUT_DEV_matching_withNN.pdf
ForwardEffpLongfromB_2_eta_5_noUT_DEV_matching_withNN.pdf ForwardGhostRate_p_noUT_DEV_matching_withNN.pdf
forward_then_matching_no_ut: efficiency forward_then_matching_no_ut: ghost
ForwardEffptLongfromB_2_eta_5_noUT_DEV_longbest_withNN.pdf ForwardGhostRate_pt_noUT_DEV_longbest_withNN.pdf
ForwardEffpLongfromB_2_eta_5_noUT_DEV_longbest_withNN.pdf ForwardGhostRate_p_noUT_DEV_longbest_withNN.pdf
NEW forward_then_matching_no_ut: efficiency NEW forward_then_matching_no_ut: ghost
ForwardEffptLongfromB_2_eta_5_noUT_withNN.pdf ForwardGhostRate_pt_noUT_withNN.pdf
ForwardEffpLongfromB_2_eta_5_noUT_withNN.pdf ForwardGhostRate_p_noUT_withNN.pdf
Edited by Alessandro Scarabotto

Merge request reports