How can i adjust a parameter if there was no improvement after x iterations?

Answered

Céline Meister

April 11, 2022 14:08

• 
• 
• 

I have the following (working) model, which i have to improve:

timeLimit=200   
numVarAct=int(0.5*n)
numIterations=100
 
mdl.params.timeLimit=30
 
 
tm=time.time()
tm2=time.time()
 
while tm2-tm<timeLimit:
    for i in V:
        for m in M_i[i]:
            x[i,m].lb=x[i,m].x
            x[i,m].ub=x[i,m].x
         
    actVar=np.random.choice(V,numVarAct,replace=False)
        
    for i in actVar:
        for m in M_i[i]:
            x[i,m].lb=0
            x[i,m].ub=1
 
    mdl.optimize()

Can anyone help me how i can increase the parameter numVarAct by 1 if the objective value doesn't increased after say 5 iterations?

I assume that it's something in the form

if ... > ... :
    numVarAct = numVarAct + 1

...but i have no idea how 1. define 5 iterations and 2. "measure" the improvement so that i can build an if - constraint. 

0

• Official comment

  

  Simranjit Kaur

  • Gurobi Staff

  November 16, 2025 23:39

  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.
• 

  Jonasz Staszek

  • Community Moderator

  April 12, 2022 06:25

  • 
  • 
  • 

  Hi Celine,

  Re 1: You can do that with help of a counter variable. You'd need to define it before your while loop, and then increase it by one every time you call mdl.optimize().

  Re 2: This can be achieved by creating a list in which you'll store the solution values of consecutive runs. You would - again - need to initialize such a list before your while loop, and then - after each optimization call - query the optimal value parameter of your model and store it in the list you created (assuming that your models solve to optimality).

  With these two at hand, you can create a condition that accounts for both the given number of iterations or the unchanging optimal solutions quality.

  Your code could look something like this:

  timeLimit=200   
  numVarAct=int(0.5*n)
  numIterations=100
   
  mdl.params.timeLimit=30
   
   
  tm=time.time()
  tm2=time.time()
  solutions = []
  counter = 0 
  while tm2-tm<timeLimit:
      for i in V:
          for m in M_i[i]:
              x[i,m].lb=x[i,m].x
              x[i,m].ub=x[i,m].x
           
      actVar=np.random.choice(V,numVarAct,replace=False)
          
      for i in actVar:
          for m in M_i[i]:
              x[i,m].lb=0
              x[i,m].ub=1
   
      mdl.optimize()
      counter += 1
      solutions.append(mdl.ObjVal)

  and the if-condition could look something like this:

  if counter == 5 and solutions[0] == solutions[1] == solutions[2] == solutions[3] == solutions[4]:
      numVarAct = numVarAct + 1

  You can obviously improve this code or adjust it to your needs.

  Hope this helps.

  Best regards
  Jonasz

  0
• 

  Céline Meister

  April 12, 2022 06:38

  • 
  • 
  • 

  thank you so much! 

  0

Post is closed for comments.