Constraint involves log

回答済み

XIAO ROU YU

• 2024年05月23日 11:10
• 編集済み

I want to use Gurobi solve a problem with log constraint

The origin CVXPY code is:

w = cp.Variable((n, 1))
Z = cp.Variable((n, N))

for j in range(N):

column_Zj=cp.reshape(Z[:, j], (-1, 1))

log_expression=-cp.log(1+ (w.T @ R_stack)) +column_Zj.T @ R_stack

constraints.append(log_expression <=-s[:,j])

And I tried to turn into Gurobi code:

w = model.addMVar((n, 1), name="w")
w_t = w.transpose()
Z = model.addMVar((n, N), name="Z")

for j in range(N):

column_Zj=Z[:, j]

column_Zj_t=column_Zj.transpose()

x=model.addVar(name="x", lb=-1)

y=model.addVar(name="y", lb=-1000)

model.addConstr(x == 1+ (w_t @ R_stack), name="log inside")

model.addGenConstrLog(x, y, name="log")

log_expression=model.addVar(name="log_expression")

model.addConstr(log_expression == -y + column_Zj_t @ R_stack)

model.addConstr(log_expression<=-s[:, j], name=f"log_expr")

model.Params.FuncNonlinear=1

model.setParam('MIPGap', 0)

 

 But when using Gurobi to solve, I got the solution :

Model is infeasible or unbounded Best objective -, best bound -, gap -

 

I think the two code above is the same, it seems that the Gurobi code is wrong

I can't find the error

関連するエンティティ

• General Constraints
• Infeasible

0

• 

  Riley Clement

  Gurobi Staff

  • 2024年05月24日 12:50

  Hi Xiao,

  By default Gurobi variables have a default lower bound of 0.

  Could this cause the issue here?

  - Riley

  0
• 

  XIAO ROU YU

  • 2024年05月25日 15:16

  

  

  Hi Riley,

  Thanks for answering.

  I tried to set the variables who may be negative to a negative lower bound,

  but still encountered an infeasible model.

  Therefore, I suspect there may be an error in the code for the log term,

  but I cannot find the error.

  0
• 

  Riley Clement

  Gurobi Staff

  • 2024年05月25日 16:02

  Hi Xiao,

  I would suggest the following article will help
  How do I determine why my model in infeasible?

  - Riley

  0

サインインしてコメントを残してください。