Thanks, Oscar! I studied it but it uses autodiff = :true (which will use auto differentiation I guess with dual numbers (I am quite bad with these numerical methods)). I just wanted to make a function so that it won’t use a dual number. I did not quite understand whether it is possible or not. See the picture… (https://jump.dev/JuMP.jl/stable/nlp/#User-defined-Functions-1) Does that mean we can define our own function and its gradient? Optim.jl, LsqFit.jl actually optimized parameter of the external simulator. (I wanted to make such routine).
