·
Probabilistic Tagging with GNNs
Weight every simulated event by the probability that its jets pass the tagger, instead of keeping only the events that pass. A graph network predicts that probability from the whole event.
Key result
Embedded
in ATLAS analyses
Used in several ATLAS analyses, including the legacy VH(bb/cc) measurement, where it enters the classifier training, the uncertainty estimates and the final result.
Resources