Skip to main content

Speed linear optimization model

Answered

Comments

2 comments

  • Maliheh Aramon
    Gurobi Staff Gurobi Staff

    Hi Claire,

    Some general recommendations you could consider experimenting with:

    • Use the latest Gurobi version 11.0.2.
    • Leave the Threads parameter at the default value
    • Consider more aggressive presolve options by setting Presolve=2 and PreSparsify=2. Does CPLEX result in a smaller presolved model?
    • Consider experimenting with barrier parameters such as BarOrder, CrossoverBasis, and Crossover.
    • Your log shows the Crossover as the bottleneck with warning messages indicating numerical issues. Consider setting CrossoverBasis=1 to start the crossover with a more numerically stable initial basis.
    • Consider setting the NumericFocus parameter to a higher value than the default value of 0.
    • Enable the quad precision computation in crossover simplex right from the beginning by setting the Quad parameter to value 1.

    Best regards,

    Maliheh

    0
  • Claire Lambriex
    First Comment
    First Question

    Hi Maliheh,

    Thanks for you answer! I will try some of the things you mentioned.

    Is there some guide from gurboi about model formulation and general problems for large LPs?

    I think some of my problems might be due to my model formulation, but I am not sure how to adjust this.

    Best regards,

    Claire

    0

Please sign in to leave a comment.