Andreas Noack
Andreas Noack
I've reduced the problem size of the benchmarks to not exceeding 10000 elements, and that goes through without hitting the memory limit. The overall runtime is slower, which shouldn't be...
@palday It would indeed be good if you could try out some of your larger examples. If we conclude that supporting large problems without uncertainty is important, I can see...
@devmotion thanks for the comments. I believe I've addressed all of them now.
@palday did you get a chance to run your examples? I'm quite eager to get error bounds in Loess plots, so it would be good to find a path forward...
@palday bump. Do you object to this or are you okay with moving forward here?
@devmotion it looks like the benchmarks now complete but then fail afterwards with an error about the repo being dirty. Have you seen that error before?
@devmotion your intuition was right here. With Julia 1.12.2, the `Pkg.dev` command added a sources line to the Project.toml. For now, I'm working around the error by committing the change...
This was an item in the list in https://github.com/JuliaStats/GLM.jl/issues/500#issue-1390221148. Not sure if it has been discussed elsewhere. I just converted it to an issue to have a place to clearly...
> I'm a bit confused what the use of the square `Q` is The way I see it is that a compact representation of `Q` naturally is square. That is...
I think we should wait for #487. It seems really close. I'll update this PR once #487 has been merged.