I’m not quite sure if I chose the right domain. However, this question is about automatic differentiation.
My plan is to overload all functions (for my own tracked real/array type) for which rules exist in ChainRules.jl.
Is there a way to do this automatically, or is there a list of all rules that ChainRules.jl contains that allows all functions to be automatically overloaded?
I believe GitHub - JuliaDiff/ChainRulesOverloadGeneration.jl: Tools to help generate operator overloads based on ChainRules was created for this purpose. GitHub - gaurav-arya/StochasticAD.jl: Research package for automatic differentiation of programs containing discrete randomness. and GitHub - invenia/Nabla.jl: A operator overloading, tape-based, reverse-mode AD use it. However, I’ve also heard it has a number of sharp edges, so if you run into any time sensitive issues you might want to ask on the Slack #autodiff channel.
Okay, thank you! Then I’ll take a look at the whole thing.