Skip to main content

Solver gets stuck in Root simplex log...

Answered

Comments

4 comments

  • Official comment
    Simranjit Kaur
    • Gurobi Staff
    This post is more than two years old. Some information may not be up to date. For current information, please check the Gurobi Documentation or Knowledge Base. If you need more help, please create a new post in the community forum, or try Gurobot, our chatbot interface offering instant, expert-level support.
  • Jonasz Staszek
    • Community Moderator
    • Gurobi-versary
    • Thought Leader
    • First Question

    Hi Patrick,

    without knowledge about the model itself, I cannot give you specific guidance.

    Nonetheless some things I would try come below:

    1) Update Gurobi to v10 - you seem to be using v. 9.5.1.

    2) Read Gurobi guide for numerical issues.

    3) The range of coefficients of your model is huge. Perhaps you could reformulate it to make this range smaller? This will most likely help with the numerics too. You can read more here.

    4) In my opinion, setting Crossover to 0 in your case won't help - you are solving a MIP and hence you need a basic solution as a start for the B&B. With Crossover set to 0, interior solution computed by barrier is returned. Read more here.

    5) It's not obvious that increasing the number of threads will make your progress faster - read here, here and here.

    All in all, the common theme in all these points are numerics of the model. You should perhaps start with these. A good first step would be to work with the coefficient ranges of your model (see point 3).

    Hope these thoughts will be of some use to you.

    Perhaps someone from Gurobi team can comment further?

    Best regards
    Jonasz

    1
  • Matthias Miltenberger
    • Gurobi Staff

    Hi Patrick!

    It seems like the simplex does not converge to a solution. As Jonasz already noted, this is most likely a numerical issue.

    You could try to avoid basic solutions altogether by setting the parameter NodeMethod to 2 and also use barrier for the node relaxations - this will lead to longer node LP solving times but at least there is hope to start the tree search at all.

    Another thing to try is the NoRel Heuristic (via NoRelHeurTime) to find a feasible solution at least.

    Cheers,
    Matthias

    0
  • Patrick Freitag
    • Gurobi-versary
    • First Comment
    • First Question

    Dear Jonasz and Matthias,

    thank you so much for your help.

    Best,

    Patrick

    0

Post is closed for comments.