Dan Foreman-Mackey
Dan Foreman-Mackey
It might be better to just add something like this to the docs rather than trying to support a feature since I expect it would be hard to do shape...
Thanks for this Brett! I think that either approach would be just fine (I'm not scared of dependency management :D) so it's totally up to you. I also agree that...
This error is raised when the Gaussian Process hyper parameters end up in a part of parameter space that is numerically unstable. This can include places where the length scale...
I have seen this before, and I don't totally understand why this happens. In the short term, I'd recommend using the `groups` argument to `to_netcdf` (care of @avivajpeyi): ```python trace.to_netcdf(...
The issue is somehow related to the fact that `observed_data` is required to be fixed (i.e. the same for every sample), but here we're overloading the `obs` argument to be...
Yes. This is exactly the same issue. The "observed data" isn't a constant if it depends on the model!
Haha good point. The celerite2 design (mistakes?) are the same and also my own :D
Thanks! If you're interested in a Julia version, there has been some work on implementing at least some parts over in: https://github.com/JuliaAstro/Transits.jl And we have this old implementation of celerite...
Pinging @iancze
Thanks @danhey! I think the relevant comparison is the current figure here: https://docs.exoplanet.codes/en/latest/tutorials/data-and-models/#orbital-conventions I would imagine that this should replace or complement the existing one?