Gurobi 9.0.1 - Non-convex quadratic optimization scalability

Answered

Zhen Lin

February 27, 2020 03:48

• 
• 
• 

I have been wondering what would be the largest size of the non-convex quadratic problem that Gurobi 9.0.1 can solve. Namely, the scalability.

For example, In Julia, I have tried Gurobi 9.0.1 with randomly generated problems with non-convex quadratic objective function and quadratic constraints. It could solve the problem in seconds when there were 5 variables. However, Gurobi became very slow when the problem had 10 or more variables. Is this expected? Just want to make sure that I'm not missing anything.

Thanks!

0

• Official comment

  

  Simranjit Kaur

  • Gurobi Staff

  June 23, 2025 03:17 Edited

  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?.
• 

  Tobias Achterberg

  • Gurobi Staff

  February 27, 2020 11:49

  This question is very hard to answer, because just like in MILP, the size is not necessarily the most important factor when it comes to the time needed to solve a model. The more important aspect is the structure of the model.

  And given that, randomly generated problems usually have a very poor structure, so it is not very surprising that you are unable to solve them. You should instead try problem instances of the actual application that you want to address.

  You may also want to take a look at the publicly available results on the QPLIB. Prof. Hans Mittelmann has some benchmark results: http://plato.asu.edu/bench.html (non-convex MIQCP is under "Mixed Integer QPs and QCPs").

  Best regards,

  Tobias

   

  0
• 

  Zhen Lin

  February 28, 2020 23:14

  • 
  • 
  • 

  Thank you Tobias!

  0

Post is closed for comments.