Parameter tuning - guided search

Answered

Grzegorz Siekaniec

• January 20, 2020 08:37

As far as I know, an automatic tuning tool perform its own search procedure through parameter space. It would be nice to be able to guide a search through a specific set of parameters and its values, like it is done in scikit learn (https://scikit-learn.org/stable/modules/grid_search.html). So for instance one might want to specify the following set of parameters to be evaluated:

params = {
    'Heuristics': [0.1, 0.2, 0.4],
    'MIPFocus': [1, 2],
    'Presolve': [0, 1, 2]
}

 

Would it be possible? Or maybe already somehow is.

0

• Official comment

  

  Simranjit Kaur

  Gurobi Staff

  • December 06, 2024 00:50

  This post is more than three 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 why not try our AI Gurobot?.
• 

  Matthias Miltenberger

  Gurobi Staff

  • May 19, 2021 10:18

  Hi Grzegorz,

  Apologies for the (very) late reply!

  You can specify custom parameter sets when starting a tuning run to kickstart the progress if you already have a good idea of what works well:

  TuneBaseSettings (gurobi.com)

  Tuning MIP models is often pretty complicated and a fully automatic process is not always providing the best results. This is due to the nature of MIP solving in regards to performance variability and the sheer amount of parameters that may have a large range of side effects.

  Please refer to our documentation about the parameter tuning tool for more information:

  Parameter Tuning Tool (gurobi.com)

  Cheers,
  Matthias

  0

Post is closed for comments.