Benjamin Pope

Thanks! Yeah I think the main issue is that there is currently a very nice API for automatically selecting which CBVs to use. It would be nice to be able...

Yes well - tackling one part of the problem at the time. For these bright stars the functionality of using nearby stars for optimization is useless, and we will need...

Gorgeous! I wonder - it looks like the pulsations are tidally induced, and so they're phase coherent with the eclipses. What does this mean for the eclipse mapping? On Fri,...

Seconded - hoping to use J1, J0 in simple interferometry applications, without necessarily needing the higher order. Is there a reason these are hard to do in Jax?

Just bumping this @jakevdp - I think the recurrence identities for higher integer Jv are not necessarily numerically stable. I think float order Jv is probably too hard basket right...

Hi @jakevdp, Are we happy to merge this? Cheers, Ben

I've also implemented spherical Bessel functions with my student @shashankdholakia, with custom vjp - if you don't want these in core jax, do we just want a separate repo with...

So a quick update on what would be needed for good performance in `float32`: ## J0, J1, Jv The Jax translation of the implementation in CEPHES used in scipy (which...

I actually see Bessel Jv has now been implemented! But it seems to be numerically unstable: ```python def j1(x): # jax version return jax.scipy.special.bessel_jn(x,v=1)[1] from scipy.special import j1 as j1s...

Just to bump this - for our use case we seem ok with a bit of a hack, patching together the core Jax bessel function for small x with a...