Closed drewoldag closed 3 months ago
I did some initial testing and I'm seeing some unexpected behavior. Then again, I'm not 100% sure of what the true behavior is either, but these all represent deviations from JPL Horizons ephemeris for 2002 PV170 for a 24-hour period beginning UT 2019-08-29. The original fits (blue dots for CTIO, orange for the geocenter) were also off, but less so.
Actually, even disabling the method and bounds, I'm still seeing the odd behavior, so something else has changed since my (old) version of correct_parallax. I will investigate...
After more testing, the method seems to be the problem, not the bounds. If I just comment out the method line and leave the bounds in, I'm able to successfully reflex-correct orbits that failed previously, like (43500) Chandler, 282P, and (6478) Gault.
Ok, good to know. Thanks for taking a look at this. Jeremy has shared an analytic solution for this with me as well. I'll try to get that in the PR tomorrow and perhaps we can spend some time running the code together. Or if you want to share the script that you used to produce the image above, that would be great too.
Check out this pull request on
See visual diffs & provide feedback on Jupyter Notebooks.
Powered by ReviewNB
This PR creates a new dispatcher function with the name
correct_parallax
that will utilize either the baseline (renamed to correct_parallax_with_minimizer) or the exact geometric algorithm.The baseline has been modified as well to allow the user to specify a particular scipy minimizer algorithm, as well as a bound on the total search space.
I've updated tests and added a demo notebook is included as well that shows how the two algorithms compare.