search

mlubin / ReverseDiffSparse.jl

Reverse-mode automatic differentiation for sparse Hessians
Other
24 stars 12 forks source link

JuMP AD #44

Closed stumarcus314 closed 7 years ago

stumarcus314 commented 7 years ago

If JuMP is used to compute the Hessian, does it use ReverseDiffSparse? If JuMP is used to only compute the gradient, does it instead use ForwardDiff? The JuMP documentation (https://jump.readthedocs.io/en/latest/nlp.html#) indicates that both packages are used to compute automatic derivatives.

mlubin commented 7 years ago

JuMP uses ReverseDiffSparse for everything except user-defined functions. See section 5 of https://mlubin.github.io/pdf/jump-sirev.pdf.

Please use discourse for questions like these in the future.