Hello. I am implementing L-shaped method (stochastic version of Benders decompostion) and have some question

回答済み

제익 유

2025年06月30日 05:41

• 

I'm struggling to add feasibility cuts to get rid of the first stage solution which makes second stage problem is infeasible

Then I know that one should use dual unbounded ray to construct a feasibility cuts.
Then how can i get dual unbounded ray in Gurobi? Thanks.

関連

• Bender's

0

• 

  Gwyneth Butera

  • Gurobi Staff

  2025年07月01日 14:44

  I asked our new AI tool Gurobot to answer your question. Here is what it replied

  To obtain dual unbounded rays in Gurobi for constructing feasibility cuts in two-stage optimization problems, you need to use the FarkasDual attribute. This attribute provides a certificate of infeasibility that can be used to construct feasibility cuts.

  Here's how to access and use dual unbounded rays in Gurobi:

  Step 1: Enable Infeasibility Information

  First, you must set the InfUnbdInfo parameter to 1 before solving your model:

  python

  import gurobipy as gp # Create environment and model env = gp.Env() model = gp.Model(env=env) # Enable infeasibility/unboundedness information model.setParam('InfUnbdInfo', 1) # Build your model... # model.optimize()

  Step 2: Access the Dual Unbounded Ray

  After solving an infeasible model, you can access the dual unbounded ray using the FarkasDual attribute:

  python

  if model.Status == gp.GRB.INFEASIBLE:     # Get the dual unbounded ray (Farkas dual)     farkas_dual = model.getAttr('FarkasDual', model.getConstrs())     farkas_proof = model.getAttr('FarkasProof')

   

  0
• 

  Riley Clement

  • Gurobi Staff

  2025年07月02日 05:26

  Note that if your subproblem is in dual form, and unbounded you will need the UnbdRay attribute.

   

  0

サインインしてコメントを残してください。