Originally posted by: TOYSED

Hello,

 

I am using C++ and having difficulty about adding and removing variable conversions to my model since

my variable number is rapidly changing.

 

That is I start with x variables float variables. 

I change them to bool with 

IloConversion conversion = IloConversion(env, X, ILOBOOL);

Then I remove this conversion. But in the next iteration I have x+y variable. Thus, if I use the same conversion 

I only convert x variable to bool.

How can I update my models in this case? Do I need to define a new conversion for each new variable?

 

Thank you...

#CPLEXOptimizers
#DecisionOptimization

You have to register a conversion for any variable you want to convert (otherwise CPLEX could not know that you want this variable converted).

Note that IloConversion is a subclass of IloExtractable. So you can store multiple conversions in an IloExtractableArray and then use IloModel::add(IloExtractableArray) and IloModel::remove(IloExtractableArray) do add and remove all conversions in one shot. That may simplify your code a little.

#CPLEXOptimizers
#DecisionOptimization

Originally posted by: TOYSED

Hello Daniel,

 

Thanks for the reply. I do understand that I need an array structure but I have some more questions.

First, if I need to define an IloExtractableArray, then how should I assign conversions to this array? Could you give me a small example...

Second, I am generating columns at each iteration and do not know how many additional columns I will have at each iteration.

So, is this a problem when defining IloExtractableArray for the first time. I guess I need a flexible structure like a vector in C++.

 

Thanks,

 

#CPLEXOptimizers
#DecisionOptimization

IloExtractableArray is extendible just like std::vector(). You may want to take a look at the reference documentation here. From that I hope it is clear how to use such an array: just use IloExtractableArray::add() to append a new element to the array.

#CPLEXOptimizers
#DecisionOptimization

Originally posted by: TOYSED

Hello,

 

I am trying to do it. But I could not achieve so far.

I have a set of variables X in the x vector.

After adding contraint and variable to the model I am defining a Bool Array (which I understand is a default Extractable Array)

IloBoolVarArray ARRAY(env, x.size());

IntegerMaster.add(ARRAY);

In this case I add x more variables to the model.

However I want to define an Array which is the bool equivalent of the original variables.

 

I also used

 

IloExtractableArray ARRAY(env, x.size(),X);

IntegerMaster.add(ARRAY);

 

But I got the error

Error    25    error C2664: 'IloExtractableArray::IloExtractableArray(const IloEnv,IloInt,const IloExtractable)' : cannot convert parameter 3 from 'IloNumVarArray' to 'const IloExtractable'  

 

What should I do? Do I need to define an IloBoolVarArray (which is a default Extractable Array what I understand from the documents) or a new Array using IloExtratableArray and convert it to Bool and/ Float regularly?

 

 

 

#CPLEXOptimizers
#DecisionOptimization

I think what you have to do is this (assuming that X is your array of boolean variables):

IloExtractableArray convertX(env);
for (IloInt i = 0; i < X.getSize(); ++i)
  convertX.add(IloConversion(env, X[i], ILOFLOAT));

After that, model.add(convertX) will relax all variables in X and model.remove(convertX) will unrelax them. If you create a new variable then just also append a new conversion to convertX.

Is that what you wanted to do?

#CPLEXOptimizers
#DecisionOptimization

Originally posted by: TOYSED

Let me summarize my algorithm again.

I have x variables at the beginning.

At each iteration I could add one or more variables.

Before running the algorithm I have to define a
IloExtractableArray convertX(env);

while(condition is satisfied){

convert X to float.

Run the algorithm.

Add new variables to the X

convert all X to bool

}

that is I need to update X regularly and at each step I need a coversion from bool to float and float to bool

Thus what you offer is true I guess. But how I will update convertX?

 

 

#CPLEXOptimizers
#DecisionOptimization

Originally posted by: TOYSED

I tried the following;

IloExtractableArray convertX(env);
for (IloInt i = 0; i < X.getSize(); ++i)
  convertX.add(IloConversion(env, X[i], ILOBOOL));

But I guess it does not work because I didn't see the integratiy constarints in the lp file.

#CPLEXOptimizers
#DecisionOptimization

I am confused now: You said your X array is an IloBoolVarArray. So initially all the variables should have type bool. Then why do you want to have a conversion that converts to ILOBOOL?. As far as I understand, what you want is a conversion to ILOFLOAT?

About extending convertX after adding a new variable: like I said, just add a new conversion for the new variable:

convertX(IloConversion(env, new_variable, ILOFLOAT));

Maybe you can show us parts of your code and tell us where things go wrong. Then it may be more clear what your problem is.

#CPLEXOptimizers
#DecisionOptimization

Originally posted by: TOYSED

Hello, 

 

The flow is as follows;   

 

    IloNumVarArray X(env,0,1, ILOFLOAT);

    IloNumVarArray AX(env,0,1, ILOFLOAT);

 

       //Convert LP MODEL TO IP MODEL

                IloModelIntegerMaster(env);

       IntegerMaster.add(Mmodel);

      

 

       IloExtractableArrayconvertX(env);

       IloExtractableArrayconvertAX(env);

 

       for(IloInti = 0; i < X.getSize(); i++)

       convertX.add(IloConversion(env, X[i], ILOBOOL));

 

       for(IloInti = 0; i < AX.getSize(); i++)

       convertAX.add(IloConversion(env, AX[i], ILOBOOL));

      

       IloCplexcplex1(IntegerMaster);  

       IntegerMaster.add(convertX);

       IntegerMaster.add(convertAX);

 

cplex1.solve();

 

 

      

       while(B > C){

                                   

                   

                          //remove conversion to obtain all float variables

                           IntegerMaster.remove(convertAX);

                           IntegerMaster.remove(convertX);

 

                                              

                           //add new variables

                                                                                                                convertX.add(IloConversion(env, X, ILOBOOL));  //it does not accept like convertX(IloConversion(env, new_variable, ILOFLOAT))

                                                           

                           convertAX.add(IloConversion(env, AX, ILOBOOL));

                                                           

 

                           IntegerMaster.add(convertX);

                           IntegerMaster.add(convertAX);

 

                           //solve with bool variables

                           cplex1.solve();

B--;

                                 

                                 

} //end while

 

When I run like this, it gives Multiple Conversion Exception.

#CPLEXOptimizers
#DecisionOptimization

This is wrong:

convertX.add(IloConversion(env, X, ILOBOOL));

(and similarly for convertAX). You are adding conversions for X again. So you will have multiple conversions for each variable in X in convertX, hence the MultipleConversionException. You should only add a conversion for the new variables. For all other variables you already have the conversion in convertX!

Like maybe so:

IloInt const oldSize = X.getSize();
// Create new variables and append them to X
...
for (IloInt i = oldSize; i < X.getSize(); ++i)
  convertX.add(IloConversion(env, X[i], ILOBOOL));

 

#CPLEXOptimizers
#DecisionOptimization

Originally posted by: TOYSED

Thank you so much Daniel,

It finally worked..:)

#CPLEXOptimizers
#DecisionOptimization