Gaurav Arya

A note: special case differentiations for the Lasso / "proximal gradient methods" (e.g. https://arxiv.org/pdf/2201.12348.pdf, https://arxiv.org/pdf/2105.15183.pdf) might pop out automatically from chain rules here, after doing an epigraph formulation of the...

That's a fair point. I had no special reason to depend on `ChainRulesTestUtils` other than specially checking AD for each backend, so I've tried to get rid of it and...

This is ready for a review!

Bump (looking for a review on this before the chain rules PRs; I'll adapt those after this has been merged)

This should be ready for another review (with https://github.com/JuliaMath/AbstractFFTs.jl/pull/69 as a dependency)

I've got `test_frule` and `test_rrule` to work for `ScaledPlan`'s too (and fixed some bugs with the rule in the process), with some minimal type piracy (see https://github.com/JuliaDiff/ChainRulesTestUtils.jl/issues/256). So things should...

Based on discussion with @stevengj, an alternate solution would be to define `AbstractFFTs.transpose` and require subtypes of `Plan` provide an implementation of it. However, on second thought I'm unsure whether...

That makes sense -- then it seems there are two options: 1) Not add any more types and ask downstream libraries to write e.g. `Base.transpose(::MyPlan, ...) = AbstractFFTs.r2r_transpose(::MyPlan, ...)`, where...

Hey, I made a draft PR here: https://github.com/JuliaMath/AbstractFFTs.jl/pull/67

(copying from Slack) I think it's because you're communicating plans between workers, which you're not allowed to do (https://discourse.julialang.org/t/fft-plan-cant-be-sent-between-processes/877). The function f that you pass to pmap uses the Tpx...