David Widmann

Since both PalmerPenguins and MLDatasets use DataDeps, I guess one could just load PalmerPenguins.jl instead of re-implementing everything from scratch? I don't plan to retire PalmerPenguins anyways since it's much...

I wonder if it would just work if one uses the same approach as in https://github.com/JuliaDiff/FiniteDifferences.jl/pull/146 - similar to `Diagonal`, `Symmetric` is just a representation of a specific lower-dimensional manifold...

Although it is a fundamental change and probably breaks some things, I think one should change `to_vec` as suggested in https://github.com/JuliaDiff/FiniteDifferences.jl/issues/132. I would expect `to_vec(x)` to return a canonical representation...

I reverted the update of the GP tutorial for now, I think it should get a major revision anyway similar to #305.

As far as I know, in principle loops should work with all AD backends. I guess your problems might rather be due to Zygote not allowing mutations?

Point 1 might be fixed in the latest iteration of the diffeq tutorial.

Hi! Your questions remind me of https://github.com/TuringLang/Turing.jl/issues/1302 :slightly_smiling_face: Do the suggestions in https://github.com/TuringLang/Turing.jl/issues/1302#issuecomment-645390404 fix your issues?

Yes, one should use `Array(group(...))` instead of `.value.data`. I assume the tutorial was written and executed with an old version of MCMCChains and never updated. Which version of Julia and...

I was wondering the same and noticed that `show(stdout, MIME"text/plain"(), suite)` displays only a limited and non-verbose summary. For a more verbose output one has to provide the `verbose` and...

> Otherwise Aqua.jl ambiguity testing is broken for all dependent packages. You can specify `recursive=false` to only check ambiguities in your packages: https://juliatesting.github.io/Aqua.jl/dev/#Aqua.test_ambiguities-Tuple{Any}