Conditional inside a function to be optimized

Specific Domains Optimization (Mathematical)
,
16

There are a few comments leading you astray in this thread, so let me summarize.

  • Your problem is nonlinear. You need to use the nonlinear interface for JuMP. In particular, you need to use a user-defined function: https://jump.dev/JuMP.jl/stable/nlp/#User-defined-Functions-1.
  • The nonlinear interface requires functions with scalar inputs, so s can’t be a vector input. Read this suggested work-around.
  • @Henrique_Becker is correct that @objective(model, Max, C_opt(s) evaluates this function with the arguments as JuMP variables before passing it to the solver. This is a little confusing, but is very helpful in some cases.
  • I don’t understand the need for OffSetArray, but it looks like your nonlinearity is just the single max(0, C)? Due to nondifferentiability and the large flat area, Ipopt tends to struggle with this. You might get a local optima.
  • There is nothing wrong with how you have defined @variable.
  • Your issue with Model(GLPK.Optimizer) is because you are using an old version of JuMP, although I think you’ve fixed this in another issue.
4 Likes