Skip to content

Adding PbPb NN weights for ghost probability with flatting issue fixed

Ivan Cambon Bouzas requested to merge jcambonb_PbPb_gprob_v2 into master

Due to the issues seeing the track ghost probability during the PbPb data-taken, it was found that the problem was related to the flattening of the trained NN model. The fix suggested was to put smoothen=False in

model.add_flatten_layer(flatten_data['output'], [0, 1], smoothen=False)

line. Now the track ghost probability obtained are the following

track_ghost_prob_distributions_PbPb_weights_v2.pdf

track_ghost_prob_distributions_nout_PbPb_weights_v2.pdf

Much closer to pp ghost probability distribution shape

This is linked to: lhcb/Moore!4047 (merged)

Edited by Ivan Cambon Bouzas

Merge request reports