Originally posted by: newuser2010

Hi, suppose there are 7000 binary variables, and from some information I can know 1500 binary variables are 1. How can I fix these variables in the preprocessing? For example, can I write something before
minimize …….
subject to {
…….
}

Thank you.
#DecisionOptimization
#OPLusingCPLEXOptimizer

Hi,

let me give you a small example

dvar int x in 0..5;
 
execute
{
 x.UB=4; 
 x.LB=4;
 // This sets x to the value 4
}  
 
maximize x;
subject to
{
 
}

And the solution will be x=4

Alex
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

Hi,

I made the experience that you might loose performace, if you
set the lower und upper bound of several thousands of variables,
as Alex described.

An alternative way to tackle the problem is to set up a subset
of variables for which you set up constraints that fix the
variable.

I.e. you will have sometrhing like (code is not tested)

range numVars = 1..5000;
dvar int x[numVars] in 0..5;
{int} fixedVars = ... add some code to determine the fixed vars ...
 
maximize sum(i in numVars ) x[i];
subject to
{
  // ... some other constraints ...
 
  // fix variables
  for(i in fixedVars )
   x[i] == 4;
}

You might try both. Both should yield the same solution.

Regards
Norbert
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: newuser2010

Hi, thanks for the reply.
The reason why I do not want to put that in the constraints is that the fewer constraints, the less CPU time.
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

Hi,

don't take that for granted. The reason why I proposed the approach
was that this resulted in a lower overall CPU-time ...

Regards
Norbert
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: newuser2010

Hi Norbert,

I wanna try that.
When I debug it, it cannot step into the loop. Would you please help me to take a look? Thank you very much.

The .mod file is:

int nbProviders = ...; // Number of suppliers
range providers = 1..nbProviders;
int nbHours = ...; // Number of hours of auction
range time = 1..nbHours;
int Lproviders = ...;
int minupproviders = ...;
int Fproviders = ...;
int mindownproviders = ...;

dvar int Takeproviderstime in 0..1;

execute {
var p,t;
for (p in providers) {
if (L[p]>=1) {

for (t in 1..L[p]) {
Take[p][t] = 1;
}
}
}

for (p in providers) {
if (F[p]>=1) {
for (t in 1..F[p]) {
Take[p][t] = 0;
}
}
}
}

minimize ......
subject to {
......
}

The .dat file is:

nbProviders = 2;
nbHours = 3;
L=2,0;
minup=3,3;
F=0,1;
mindown=3,2;
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

Cheers,

the code is not very readable. Would you please put the
code into bracketing
{code} some code {code}
statements.
What do you mean by "cannot step into the loop"?
Is there an error message?

Two points that I can see at the moment:

for (t in 1..L[p]) {

is not a valid syntax. It should be something like

for (var t=1; t<=L[p]; t++) {

Furthermore you assign some value to a decision variable:

Take[p][t] = 1;

This is invalid.
Too what purpose?
In case you want to fix the variable you should do this as
described in this thread.
Hope this will help.

Regards
Norbert
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: newuser2010

Thank you, Norbert.
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: UDOPS

HI, I am also interested in fixing variables, primarily not to speed solution, but to reduce memory consumption. I have a large model that solves quickly but exceeds memory frequently. I can decompose the model into steps and fix the variables in later steps. Will either of these suggested methods reduce the memory consumption relative to the same model with all variables active? Or will these methods consume as much or more memory than the same model unfixed.

The alternative I suppose is for me to create alternative methods to subtract the fixed variable set from the active variable set and insert fixed values into the constraints to represent the fixed variable values.
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: UDOPS

Update: I tried it, and saw no reduction in memory. Looks like I actually have to remove the variables dynamically from the formulation.
#DecisionOptimization
#OPLusingCPLEXOptimizer