Removing redundant zero-valued variables makes the model slower to solve

Answered

Vincent

November 13, 2025 12:42 Edited

• 
• 
• 

Sometimes, when redundant variables are removed — those that take a value of 0 at optimality — the overall solve time increases. The reduced model has fewer variables, but the solver takes longer to reach optimality. Any insight on why removing such variables might make the solve harder would be appreciated. 

Related to

• Performance

0

• 

  Hamideh Anjomshoa

  • Gurobi Staff

  November 11, 2025 03:55

  To fairly compare the performance of your two models, it’s really important to use multiple random seeds. Running the model just once with a single seed doesn’t give a reliable picture of how they actually perform. I’d recommend checking out the following article and running your models with several different seeds before drawing any conclusions about their performance.
  https://support.gurobi.com/hc/en-us/articles/26992650628369-How-can-I-make-accurate-comparisons

  1
• 

  Vincent

  November 13, 2025 12:41

  • 
  • 
  • 

  Thank you, Hamideh! 

  0

Please sign in to leave a comment.