Re: Non-convex non-linear optimization problems

Answered

Dhiraj Yadav

• May 01, 2024 06:53

I have one query regarding the use of softmax function into my gurobi model. For example, I want to add constraint like

Pij[i, j] = exp(-theta[i]*T[i,j] + beta_dash[i,j]*V[i])/(sum(exp(-theta[l]*T[l,j] + beta_dash[l,j]*V[l]) for l in range(0,n)) for i in range(0,n) for j in range(0,m) 

into my gurobi model. The question is how do i do that?

Related to

• Nonlinear

0

• 

  Ronald van der Velden

  Gurobi Staff

  • May 01, 2024 07:59

  Hi Dhiraj,

  The API for non-linear functions is grouped under "General constraints" in our documentation. For composite functions like the one you describe, you would add auxiliary variables for intermediate results, e.g. have a variable to represent -theta[i]*T[i,j]+beta_dash[i,j]*V[i] and then another one to calculate the exp() of that first auxiliary variable.

  Kind regards,
  Ronald

  0
• 

  Dhiraj Yadav

  • May 01, 2024 11:16

  

  

  Thanks Ronald van der Velden. It was quite helpful and now I am able to solve the issue.

  Best,

  Dhiraj

  0

Please sign in to leave a comment.