Skip to content

Latest commit

 

History

History
19 lines (14 loc) · 1.33 KB

model-handling.md

File metadata and controls

19 lines (14 loc) · 1.33 KB

Discussion notes model handling

Wednesday, Nov 25th 16h30

What to do with the BackgroundModel?

  1. Remove BackgroundModel and handle the case with SkyModel(spatial_model=TemplateSpatialModel(), apply_irf={"edisp": False, "psf": False, "exposure": False}) The problem is, that the latter case is possibly anyway, but untestedt and it's unclear whether it already works.
  2. Rename to e.g. DataModel, but then the case SkyModel(spatial_model=TemplateSpatialModel(), apply_irf={"edisp": False, "psf": False, "exposure": False})m should probably be forbidden...

Model bookeeping in .to_image(), .to_spectrum_dataset() etc.

  • Right now we rely on .npred_background() in MapDataset.to_image(), MapDataset.to_spectrum_dataset() etc, This is non ideal, as it requires redefining and re-setting model parameters, e.g. https://github.com/gammapy/gammapy/blob/master/gammapy/estimators/flux.py#L164
  • There are a few possible options:
    • Rely on .background instead and reset the model as a whole
    • Rely on .background and copy (?) over the model, while adpating the dataset name for the FoVBackgroundModel
    • Rely on .npred_background and reset the FoVBackgroundModel?

Add minimal model management?

  • Methods like Model.select_region(), Models.to_template_spatial_model(geom=), could be useful and are simple to implement...