Approximation of nonlinear terms

Answered

Yunhui Lin

• October 20, 2022 06:31

Dear support team,

I am using the general constraint function of Gurobi to model nonlinear terms, such as log-function by "addGenConstrLog". Here, we can specify the approximation errors. It works well, and for some problems, Gurobi's performs rather impressive.

However, I am not sure how Gurobi does the linearization. Is it by inner piecewise linearization or outer piecewise linearization, or a combination of both? In more general cases, what is the rule applied (i.e., approximate a function from below or above)? 

Thank you very much!

Infi

0

• 

  Jaromił Najman

  Gurobi Staff

  • October 20, 2022 08:42

  Hi Infi,

  Gurobi usually chooses points on the original function to construct the piecewise-linear approximation. This behavior can be controlled via the FuncPieceRatio parameter.

  Best regards, 
  Jaromił

  0

Please sign in to leave a comment.