Model infeasible for with general constraint genConsrSin

回答済み

Zhizhen Qin

2023年05月03日 05:52

• 
• 
• 

I have trying to use Gurobi to minimize non convex function Levy (https://www.sfu.ca/~ssurjano/levy.html), and I couldn't get the optimal solution. After further digging, I found the following IIS when I manually tried to limit the solution space to around optimality:

\ Model _copy
\ LP format - for model browsing. Use MPS format to capture full model detail.
Minimize
  0
Subject To
 w0: - 0.25 x0 + w0 = 0.75
 sin_1_in: - 3.141592653589793 w0 + sin_1_in = 0
 const_sin_inter_in0: - 3.141592653589793 w0 + var_sin_inter_in0 = 1
 x0upper: x0 <= 4
Bounds
 w0 free
 sin_1_in free
 var_sin_inter_in0 free
General Constraints
 sin_1_out: sin_1_out = SIN ( sin_1_in )
 const_sin_inter_out0: var_sin_inter_out0 = SIN ( var_sin_inter_in0 )
End

Solving this model with gurobi_cl would result in infeasibility. However, I cannot deduce anything conflict when I try writing out the above constraints by hand. May I get any help on why the model would be infeasible?

 

Many thanks!

関連

• Infeasible
• General Constraints

0

• 

  David Torres Sanchez

  • Gurobi Staff

  2023年05月03日 14:28 編集済み

  Hi Zhizhen,

  The error comes from the bounds of the resulting sin variables, the ones containing \(\texttt{out}\).
  The constraints imply these variables must be negative, but the default LB of a variable is 0 (e.g. see model.addVar).
  If you explicitly set the lower bound to -1 and the upper bound to 1, for these variables, the problem should be feasible.

  Cheers, 
  David

  1
• 

  Zhizhen Qin

  2023年05月04日 00:34

  • 
  • 
  • 

  Hi David:

   

  The problem indeed is due to the range of variables. After setting the variables to correct ranges I am now able to find a feasible solution to the posted instance, and to minimize the Levy function which is my original task. Thanks a lot!

  
  Best,

  Zhizhen

  0

サインインしてコメントを残してください。