Can SCIP solve MINLP with trigonometric functions?

Specific Domains Optimization (Mathematical)
1

Hi everyone!

I’m trying out SCIP in JuMP (with SCIP.jl). In the FAQ in the documentation, when aswering the question “Which kind of MINLPs are supported by SCIP?” (SCIP Doxygen Documentation: Frequently Asked Questions (FAQ)), it is written that SCIP does not support trigonometric functions. However when I wanted to try it out, the solver did manage to solve a problem with trigonometric function. The problem is:

import JuMP, SCIP

model = Model(SCIP.Optimizer)
set_attribute(model, "display/verblevel", 0)
set_attribute(model, "limits/gap", 0.05)

@variable(model, 0<= x <= pi)
@variable(model, y)

@objective(model, Min, y)

@NLconstraint(model, c, y>= sin(x)*sin(11*x))

optimize!(model)

Was SCIP updated to support trigonometric functions but not the documentation? Thanks!

(I’m using JuMP v1.22.2 and SCIP v0.11.6)

1 Like
2

SCIP supports sin and cos.

See GitHub - scipopt/SCIP.jl: Julia interface to SCIP solver

In general, SCIP will throw an error if you try to use an unsupported function.

1 Like
3

Thanks for the answer!

Do you know if trigonometric functions are natively supported by SCIP or if it is the Julia wrapper that reformulates the functions in a way the solver can handle ?

4

SCIP supports sin and cos. The Julia wrappers do not reformulate approximations of nonlinear expressions; we pass them directly to the solver.

5

Thanks for the feedback. I’ve created a SCIP issue: Documentation update about trigonometric functions (cos, sin) support · Issue #98 · scipopt/scip · GitHub

6

Feedback from SCIP devs: support was introduced in SCIP 8.0.0 (early 2022), see: