Skip to main content

Same box-constrained nonconvex QP: direct x’Qx formulation solves in 397 s, explicit lifting X = xx’ stalls at 11.8% gap after 7200 s. Why?

Answered

Comments

1 comment

  • Official comment
    Hassan Hijazi
    • Gurobi Staff

    Gurobi's internal lifting of bilinear terms appearing in the objective will try to detect variables that would only take 0 or 1 values at optimality, this is no longer triggered if a manual lifting is applied. 
    My recommendation is to export the presolved model for your research purposes using (e.g., in python):
    reduced_model = model.presolve()
    reduced_model.write("reduced_model.lp")
    and applying your cuts on the presolved model.

Please sign in to leave a comment.