Cite DifferentiationInterface and ImplicitDifferentiation?

[gdalle]

Hi and congrats on the huge review, which will be tremendously useful to everyone in the field. I also really like your fully open science approach, and will emulate it in the future.

Since your paper does contain a broad overview of autodiff and implicit differentiation in Julia, along with the more specific stuff on ODEs, I wanted to ask if it would make sense to add citations for two more packages. Spoiler alert: I created both of them, so of course I'm biased and I won't be offended if you think they are irrelevant.

1. ImplicitDifferentiation.jl seems related because it is a more generic (albeit less performant) approach to implicit gradients in Julia. In the SciML ecosystem, the implicit function theorem is implemented under the hood, and users don't need to provide the optimality conditions for, say, an optimization problem or a nonlinear solve. In ImplicitDifferentiation.jl, users have to write down the equation to which said theorem is applied. Granted, it doesn't apply well to ODEs, but you cite other parts of the Julia ecosystem that are even more remote (like Diffractor.jl). Ping @mohamed82008 (the co-creator).

2. DifferentiationInterface.jl is the successor to AbstractDifferentiation.jl, which is also cited in the paper. Unlike AbstractDifferentiation.jl, it is gaining rapid adoption in the Julia ecosystem, first and foremost inside SciML packages. It also contains a brand new mechanism for sparse autodiff, which we hope will be able to replace SparseDiffTools.jl. Ping @adrhill (the co-creator).

What do you think?

[facusapienza21]

Hi @gdalle ! Thank you for your suggestion. I am glad you were able to post this just before we published the pre-print :)

I had included the reference to DifferentiationInterface.jl in the manuscript as we found it appropriate. Thank you so much.