Skip to main content

How to use non-defaults solution algorithm in Gurobi?

Answered

Comments

3 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.
  • Maliheh Aramon
    • Gurobi Staff

    Hi Tanmoy, 

    The Gurobi Optimizer uses an LP-based Branch-and-Bound (B&B) algorithm to solve a MIP. The B&B algorithm has multiple building blocks including presolve, relaxation, node selection, branching, cutting planes, heuristics, and conflict analysis. The main idea of B&B is to keep track of two bounds and close their gap as fast as possible. 

    In a minimization problem, an upper bound represents the objective value of the best feasible solution found so far and a lower bound represents the best objective value that can be possibly found. We refer to the former as incumbent and to the latter as best bound. 

    Gurobi uses its different building blocks to expedite the process of finding good bounds and closing their gap.

    As a user of Gurobi, you can potentially control how fast a good incumbent is found via 1) setting parameters such as Heuristics, NoRelHeurTime, MIPFocus=1, and RINS, for example, 2) providing a start solution, 3) providing custom heuristic solutions via callbacks. 

    You can also potentially control how fast a good bound is found via 1) setting parameters such as Presolve=2, Cuts=2|3, MIPFocus=2|3, VarBranch=3 for example, and 2) providing valid user cuts via callbacks. 

    You might check our documentation on MIP - A Primer on the Basics for an introductory description of how B&B works.

    Best regards,

    Maliheh

    1
  • Tanmoy Das
    • Gurobi-versary
    • Investigator
    • Collaborator

    Appreciate it. I will aim for that.

    0

Post is closed for comments.