lsst.ip.diffim
16.0-4-gcfd1396+5
|
1. evaluate necessity for separate parameters for pos- and neg- images
2. only fit background OUTSIDE footprint (DONE) and dipole params INSIDE footprint (NOT DONE)?
3. correct normalization of least-squares weights based on variance planes
4. account for PSFs that vary across the exposures (should be happening by default?)
5. correctly account for NA/masks (i.e., ignore!)
6. better exception handling in the plugin
7. better classification of dipoles (e.g. by comparing chi2 fit vs. monopole?)
8. (DONE) Initial fast estimate of background gradient(s) params – perhaps using numpy.lstsq
9. (NOT NEEDED - see (2)) Initial fast test whether a background gradient needs to be fit
10. (DONE) better initial estimate for flux when there's a strong gradient
11. (DONE) requires a new package lmfit
– investiate others? (astropy/scipy/iminuit?)