Originally posted by: AndyHam

Dear IBM,
I am trying to generate 10 different test instances, but the following code generates the same instance 10 times.
Please help me to find my mistake in the code.

I looked into the previous post, but it does not fix my issue.
https://www.ibm.com/developerworks/community/forums/html/topic?id=77777777-0000-0000-0000-000014401109

Thanks,
Andy

[1012,1010,1005,1021,2033,2042,2030,2003,3010,3048,3000,3012,4018,4037,4000,4044,5038,5035,5046,5032,]
[1012,1010,1005,1021,2033,2042,2030,2003,3010,3048,3000,3012,4018,4037,4000,4044,5038,5035,5046,5032,]
[1012,1010,1005,1021,2033,2042,2030,2003,3010,3048,3000,3012,4018,4037,4000,4044,5038,5035,5046,5032,]

...............

using CP;
int nj= 5;
int nv= 2; 
tuple t_Job {
key int id;
    int x;
    int y; 
    int de;    
};
{t_Job} Job;

execute { 
var now = new Date();
var mySeed = Opl.srand(now.getSeconds());
var d;
for(var i=1;i<=nj;i++) {
    d=1+Opl.rand(5);
    Job.add(i,     Opl.rand(51),Opl.rand(51), d);
    Job.add(i+nj,Opl.rand(51),Opl.rand(51),-d);    
    }
for(var v=1;v<=2*nv;v++) 
    Job.add(nj*2+v,25, 25, 0);                  
};

main {
for(var m=1;m<=10;m++){
    thisOplModel.generate();
    cp.solve();
    write("[");
    for (var j in thisOplModel.Job) 
    if(j.id<=2*thisOplModel.nj)
        if(j.id<=thisOplModel.nj)
          write( (1000*j.id+j.x) +","+ (1000*j.id+j.y) +",");
        else
          write( (1000*(j.id-thisOplModel.nj)+j.x) +","+ (1000*(j.id-thisOplModel.nj)+j.y) +"," );
     writeln("]");    
}

}                     

Hi,

 using CP;
int nj= 5;
int nv= 2;
tuple t_Job {
key int id;
    int x;
    int y;
    int de;    
};
{t_Job} Job;

execute {
var now = new Date();
var mySeed = Opl.srand(now.getMilliseconds());
var d;
for(var i=1;i<=nj;i++) {
    d=1+Opl.rand(5);
    Job.add(i,     Opl.rand(51),Opl.rand(51), d);
    Job.add(i+nj,Opl.rand(51),Opl.rand(51),-d);    
    }
for(var v=1;v<=2*nv;v++)
    Job.add(nj*2+v,25, 25, 0);                  
};

main {

 

for(var m=1;m<=10;m++){
    var opl = new IloOplModel(thisOplModel.modelDefinition);

    opl.generate();
    cp.solve();
    write("[");
    for (var j in opl.Job)
    if(j.id<=2*opl.nj)
        if(j.id<=opl.nj)
          write( (1000*j.id+j.x) +","+ (1000*j.id+j.y) +",");
        else
          write( (1000*(j.id-opl.nj)+j.x) +","+ (1000*(j.id-opl.nj)+j.y) +"," );
     writeln("]");    
}

}                     

gives

[1047,1034,1043,1025,2033,2042,2023,2004,3036,3020,3040,3026,4028,4004,4040,4010,5000,5045,5009,5036,]
[1027,1048,1042,1047,2025,2047,2013,2008,3023,3039,3024,3002,4036,4049,4050,4016,5039,5049,5042,5047,]
[1009,1024,1001,1006,2024,2020,2003,2023,3013,3040,3031,3044,4047,4015,4011,4005,5014,5011,5035,5000,]
[1015,1003,1048,1037,2018,2036,2047,2001,3022,3033,3001,3026,4024,4049,4036,4046,5014,5035,5004,5001,]
[1002,1035,1036,1031,2031,2029,2024,2042,3014,3006,3034,3000,4047,4035,4026,4009,5033,5038,5044,5021,]
[1038,1012,1020,1023,2044,2019,2048,2028,3005,3027,3010,3017,4012,4014,4009,4016,5047,5037,5030,5037,]
[1025,1036,1019,1000,2028,2022,2027,2006,3037,3000,3050,3042,4012,4000,4038,4009,5011,5005,5028,5027,]
[1015,1018,1008,1048,2044,2017,2007,2048,3032,3025,3032,3014,4035,4037,4026,4023,5031,5008,5016,5043,]
[1041,1038,1012,1000,2034,2032,2037,2003,3010,3016,3031,3048,4040,4048,4034,4006,5048,5011,5050,5038,]
[1038,1050,1049,1042,2035,2038,2003,2005,3036,3029,3031,3021,4018,4014,4035,4022,5036,5015,5044,5019,]

NB:

I used milliseconds instead of seconds

I changed your main.

regards

Originally posted by: AndyHam

Thanks! It works well now. Here is what I am trying to accomplish...
1. The script above (after I add DARP code) is to generate the training dataset for machine learning (multi-output classification).
2. The training dataset will be used to predict the allocation decision (vehicle to order). 
3. Then, the predicted values will be used as warm-start for DARP.
Eventually, I want to explore how to predict the sequencing decision, but I have a very limited knowledge of ML. If you can give some direction, it will be great!

 

Hi,

IBM Watson Studio is a very good platform for mixing ML and CPLEX

See https://developer.ibm.com/videos/introduction-to-machine-learning-and-artificial-intelligence-watson-studio-tutorial-part-1/

regards