Infeasible solution after addGenConstrPWL

回答済み

Jiajie Zhang

2022年05月28日 02:30

• 
• 
• 

Hi, I am working on the following entropy maximization problem

I use addGenConstrPWL to piecewise linearize x*(1-lnx), following are my codes to solve the model.

# generate points for PWL
ysh = zsh = np.linspace(0,1,num = 100)
hzs = [p*(1 - log(p)) if p != 0 else 0 for p in zsh]
hys = [p*(1 - log(p)) if p != 0 else 0 for p in ysh]

#build the model
msp = gb.Model('sp')
y = msp.addVars(n,m, ub = 1, name = 'y')
hy = msp.addVars(n,m,lb = -GRB.INFINITY, name = 'hy')
z = msp.addVars(n, ub = 1, name = 'z')
hz = msp.addVars(n,lb = -GRB.INFINITY, name = 'hz')
obj = gb.quicksum(q[i]*hy[i,j] for j in range(m) for i in range(n)) + gb.quicksum(q[i]*hz[i] for i in range(n))
msp.setObjective(obj,GRB.MAXIMIZE)
for i in range(n):
    msp.addGenConstrPWL(z[i], hz[i], zsh, hzs)
    for j in range(m):
        msp.addGenConstrPWL(y[i,j], hy[i,j], ysh, hys)
msp.addConstrs(gb.quicksum(y[i,j] for j in range(m)) + z[i] == 1 for i in range(n))
msp.addConstr(gb.quicksum(q[i]*c[i][j]*y[i,j] for j in range(m) for i in range(n)) + gb.quicksum(q[i]*t[i]*z[i] for i in range(n)) <= B) 
msp.optimize()

I solve the above model to get the objective value, which is very close to the true optimum. However, when I can check the solution (y,z) I find them infeasible, because the summation of them is greater than 1, e.g., 1.58, 2.36, etc.

 

Can anyone help me with this? Why infeasible solution occurs? How to get the right solution?

0

• 正式なコメント

  

  Simranjit Kaur

  • Gurobi Staff

  2025年11月16日 23:36

  This post is more than three years old. Some information may not be up to date. For current information, please check the Gurobi Documentation or Knowledge Base. If you need more help, please create a new post in the community forum, or try Gurobot, our chatbot interface offering instant, expert-level support.
• 

  Jaromił Najman

  • Gurobi Staff

  2022年05月30日 09:10 編集済み

  Your example is not reproducible. Variables n,m, q, c are not defined. In addition, there are imported packages missing. Can you please update your code to be reproducible.

  I solve the above model to get the objective value, which is very close to the true optimum. However, when I can check the solution (y,z) I find them infeasible, because the summation of them is greater than 1, e.g., 1.58, 2.36, etc.

  Do you get any warning about violated constraints in the log file/output?

  Best regards, 
  Jaromił

  0
• 

  Jiajie Zhang

  2022年06月03日 12:57

  • 
  • 
  • 

  Hi Jaromił, I did not remember the specific parameter setting. These days I have tried many other parameter settings and have not found any infeasible solutions again. In case I find any infeasible solution in my following experiments, I will post it here. Thank you very much.

   

  0

投稿コメントは受け付けていません。