search

rodluger / Limbdark.jl

Analytical transit light curves for limb darkened stars
MIT License
14 stars 4 forks source link

Compare time integration in limbdark and starry #78

Open ericagol opened 5 years ago

ericagol commented 5 years ago
ianw89 commented 5 years ago

@ericagol this might be something I could look at - would you like me to, or is this already being worked on?

ericagol commented 5 years ago

I don't think @rodluger is looking at this, and I have not yet, so I would like if you want to work on this.

ianw89 commented 5 years ago

Okay, sounds good to me!

ianw89 commented 5 years ago

@ericagol , which function is the right one to be looking at for this? integrate_lightcurve.jl?

ericagol commented 5 years ago

@ianw89 Yes, and I have an example notebook of how to call this in the notebooks/ directory.

ianw89 commented 5 years ago

@ericagol - status update on this! Firstly, I just saw your Skype messages from last Wednesday today. RIP.

Anyway, I've made progress in understanding the limbdark paper and methods. I ended up watching a bunch of khan academy videos, which helped a lot. I examined the code in tex/figures/* to see how the plots in the paper were built, and started adjust the first draft of the code I had to follow this a more.

I ran into some trouble running some of that existing code (compare_to_batman.py), which I'm still working through - the issues have to do with my machine setup (installing the pytransit package keeps failing...). I'll figure it out eventually, or just skip that bit when running locally.

There is still some open questions in my mind regarding which parameters to vary. Basically, what do you think the x-parameter of the graph(s) should be? We'd like to compare the execution time across some set of the parameter space, but which set is most interesting in your opinion? Right now I'm doing "Number of Data Points", which in practice I'm varying by changing the orbital period since that is the cleanest way to do so.

With regards to your Skype message and what library to compare it to, I will try using exoplanet until starry has the time integration code ready. Right now I'm still trying to get a plot of reasonable results working just using limbdark.

I hope this isn't blocking anything on your end - if so, please let me know!

ericagol commented 5 years ago

@ianw89 That sounds great. Yes, I think varying the number of data points is a good way to compare. Sometimes if there are few points in the light curve, the overhead will cause the time to grow sub-linearly at first, and then linearly when it is larger. You should make the grids for the two codes have similar values.