search

annayqho / TheCannon

a data-driven method for determining stellar parameters and abundances from stellar spectra
MIT License
39 stars 16 forks source link

scaling of labels #54

Open mkness opened 9 years ago

mkness commented 9 years ago

For stability we should have labels pivoting around the same order of magnitude. Ideally, all labels would be an object with a scaling (factor) and offset value. Currently there is an offset only, so values vary by a factor of ~ 100+ (i.e. the Teff scale is on the order of hundreds, the logg, [Fe/H] scale is on the order of ~1. This is not too bad but should be fixed for the general label case.

davidwhogg commented 9 years ago

see also https://github.com/mkness/TheCannon/issues/48

davidwhogg commented 8 years ago

This problem is now critical. Just sayin

davidwhogg commented 8 years ago

This issue can be closed when Eilers pull requests and you merge it.