I got up early and worked on an integrator to integrate orbits in a simplified one-dimensional dynamics problem designed to mimic the vertical properties of the Milky Way disk. I was integrating the orbits around in circles (closed orbits in 1-d kinematic space, or 2-d phase space) to get the actions and angles, but I realized that I only really need to integrate one quadrant: There are symmetries! So I switched to make use of those symmetries.
I am also working towards code that doesn't require one integral per star; I want to integrate a look-up table and interpolate to the stars. I need everything to be fast, because we have to compute these actions and angles inside the inference (optimization or MCMC) loop.
Of course part of me wonders why I am doing this at all since both galpy and gala probably do everything I need right out of the box. I just can't stop myself! (And I'm learning a lot.)
No comments:
Post a Comment