My day started with a call with Christina Eilers (MIT) about extending the spectroscopic parallax work we did with APOGEE DR14 and ESA Gaia to APOGEE DR16. This is a non-trivial extension because DR16 includes (for the first time) data from two different telescopes and this different detailed calibration, LSF, and so on. Eilers's first attempts at this seem very successful; with minimal changes she is getting comparable precision to what we got with the simpler single-telescope data. That's promising, because we didn't expect this to be easy.
She also showed that it works less well to train the model on one telescope and apply the model on the other. So something about the joint training is working. And yes, we test using held-out validation data.
No comments:
Post a Comment