Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: jointcal
-
Labels:
-
Story Points:6
-
Epic Link:
-
Sprint:AP F18-2
-
Team:Alert Production
Description
To help get a stable photometric solution, we can try implementing a magnitude-space fitter.
We should do it by creating a new MagnitudeTransform class, and related Mapping, Model, and PhotometryFitter, that all will use the magnitude instead of flux, so we still have the flux-based model available.
This ticket is to implement and test the Simple model.
Attachments
Issue Links
- has to be done before
-
- Done
Activity
The updates to DMTN-036 are probably correct, but I find the notation a bit confusing, because you're reusing a lot of symbol definitions from an earlier section that only talks about positions and fluxes, and it's not clear in this section which of the {{f}}s and {{F}}s (in particular) are still in flux units and which are now in magnitude units. In the long term (not for this ticket), I think it'd be good to rewrite with different symbols for at least model parameters in each of the three cases.
Starting in on the code now.
Looks good. I've made a number of small comments that should be addressed (but should be easy to address).
post review and rebase jenkins run: https://ci.lsst.codes/blue/organizations/jenkins/stack-os-matrix/detail/stack-os-matrix/28257/pipeline
Thanks for the quick review. We can talk sometime about how to rework the notation in DMTN-36 to make it less confusing. I agree that's for a different ticket though.
Oh, and for the record (though I was never really concerned about the SimplePhotometryModel before), the new SimpleMagnitudeModel did run to convergence on validation_data_hsc I-band, so besides the new unittest, that's at least reassuring.