Plans for starry development

[dfm]

This thread can be a place to brainstorm plans for getting sufficient support of starry-like features into JAX. I've also started a project board to track specific tasks, but this is a good place to brainstorm things at a higher level.

[shashankdholakia]

It's great to see the core functionality of starry being implemented into jaxoplanet and I'm excited to see how this goes! I wonder what is the planned scope of starry-like features in jaxoplanet. I can think of tons of useful features of starry (in no particular order):

• A starry-like API with a Primary and Secondary objects in a System (using KeplerianOrbit internally)

• Several kinds of maps:

  • Reflected light maps
  • Radial velocity maps
  • Doppler maps
  • Oblate maps (already addressed in Implement support for oblate stars in starry extension #20)
  • multiwavelength maps
  • Perhaps a nice API for loading in maps from images (like the starry map.load using a spherical harmonic transform)

Of these, I think the first point is something I'd be most excited to see. Are these features all within the scope of future jaxoplanet?

[ssagynbayeva]

Would it be possible to create light curves with jaxoplanet the way starry does? I assume that jaxoplanet would also use the starry algorithm for limb darkening (?) -- so, as Shashank mentioned, it would be awesome to combine KeplerianOrbit with the starry Kepler algorithm.

[dfm]

@shashankdholakia — Thanks for this list! I think this is all within scope, but it depends on everyone's motivation/capacity (my motivation is high, but capacity is extremely limited) to get to it. Some comments:

A starry-like API with a Primary and Secondary objects in a System (using KeplerianOrbit internally)

This is nominally already approximately supported by the existing orbit interfaces such that it shouldn't be too hard to get it running once we have a working implementation of all the starry matrices.

Several kinds of maps

This is great - can you open a new issue for each of these with a little bit more info (paper and code references, for example) so that we can track development more easily.

Perhaps a nice API for loading in maps from images (like the starry map.load using a spherical harmonic transform)

I don't think there's much that's starry-specific in the implementation in starry so that shouldn't be too hard to implement - open an issue or PR!

@ssagynbayeva — The goal isn't necessarily for jaxoplanet to reach feature parity with starry, but the hope is to be able to produce starry-like light curves in the future. We're not there yet though so using starry directly is still probably the right call for any research projects.

[ssagynbayeva]

I just want to emphasize the importance of additional starry features like starry_process (which is its own separate package but connected to starry). It makes it easier to analyze surfaces of stars using multiple light curves with a tractable Gaussian Process kernel (also makes it super easy to create a simulated light curves sampled from GP)! Though JAXifying starry_process might not make the sampling faster, but I am sure there’s some workaround (i.e. some re-parametrization) that could help the problem! Rodrigo has a great post on it. I know it’s not an urgent goal, but I felt it’s important to point this out!

[shashankdholakia]

Seconded! Interpretable GPs with spherical harmonics would be very useful for modeling starspots and surface features and adding to the starry/jaxoplanet ecosystem!