Open JBernete opened 5 months ago
Can you define "angular error"? It would be useful to explain the use case for this, and what it does in the description. I suspect the name "angular error" may be too generic - there are many angles we reconstruct, which one is this referring to?
This PR also seems to include #2497, maybe it should be parented to that branch?
With angular error I mean the angular difference between the true direction and the reconstructed direction of an event. This is useful to define event types in terms of the quality of the reconstruction.
This PR also seems to include https://github.com/cta-observatory/ctapipe/pull/2497, maybe it should be parented to that branch?
This PR needs #2497 to create the input features for the machine learning. I think we should first merge 2497 and then update here, not make 2497 bigger by including this branch there.
With angular error I mean the angular difference between the true direction and the reconstructed direction of an event. This is useful to define event types in terms of the quality of the reconstruction.
I suggest naming it "DirectionReconstructionError" or something more clear, since AngularError could refer to other angles. Or maybe better yet "DirectionReconstructionUncertainty" to avoid the class being confused with RuntimeErrors...
Or just "DirectionUncertaintly" if you want something shorter...
Adding tools to reconstruct the angular error at the DL2 level. Which will allow the definition of event types based on the angular reconstruction quality.