Right now it is already possible to analyze diffuse data, but it is a bit annoying to set up the training.
From the modeling configuration files (either regressor.yaml or classifier.yaml) it is possible to define selection cuts for training data in the section SigFiducialCuts (and BkgFiducialCuts for the classifier).
One of these cuts can be for example
offset < 1.0 # deg
but in the case of a diffuse analysis, this is not sufficient and one wants to differentiate between different offset binnings (concentric rings around a location or similar selections).
protopipe.scripts.build_models should be able to accept an array of values from the configuration file that will act like bin edges and run itself a number of times equal to the bin numbers.
The model outputs should then be written bin-wise in separate folders and protopipe.scripts.write_dl2 should then select the right model depending on the event-candidate offset.
Right now it is already possible to analyze diffuse data, but it is a bit annoying to set up the training.
From the modeling configuration files (either
regressor.yaml
orclassifier.yaml
) it is possible to define selection cuts for training data in the sectionSigFiducialCuts
(andBkgFiducialCuts
for the classifier).One of these cuts can be for example
offset < 1.0 # deg
but in the case of a diffuse analysis, this is not sufficient and one wants to differentiate between different offset binnings (concentric rings around a location or similar selections).
protopipe.scripts.build_models
should be able to accept an array of values from the configuration file that will act like bin edges and run itself a number of times equal to the bin numbers.The model outputs should then be written bin-wise in separate folders and
protopipe.scripts.write_dl2
should then select the right model depending on the event-candidate offset.