I have a canonical problem:

min c'x
s.t. Ax <= b

It's a large problem using a sparse matrix. 
What's the easiest way to formulate it in cplex? Since it's already in canonical form, I was hoping for a low level function that can accept these vars instead of using the modeling interface.

------------------------------
Zohar Levi
------------------------------

#DecisionOptimization

You may have a look at the docplex Python package. 

It provides some ways to create a model from a sparse matrix (numpy)
See http://docs.cplex.org.cn/mp/docplex.mp.advmodel.html?highlight=advmodel and the methods matrix_constraints.

You can also use a transformers from docplex: https://ibmdecisionoptimization.github.io/docplex-doc/2.21.207/mp/docplex.mp.sktrans.transformers.html?highlight=cplextransformer#docplex.mp.sktrans.transformers.CplexTransformerBase
and an example here: https://stackoverflow.com/questions/50714397/cplex-with-python-api-how-to-make-model-formulating-faster/50716526#50716526

------------------------------
Vincent Beraudier
------------------------------

Original Message:
Sent: Mon August 15, 2022 06:41 PM
From: Zohar Levi
Subject: Solve a linear programming in canonical form

I have a canonical problem:

min c'x
s.t. Ax <= b

It's a large problem using a sparse matrix. 
What's the easiest way to formulate it in cplex? Since it's already in canonical form, I was hoping for a low level function that can accept these vars instead of using the modeling interface.

------------------------------
Zohar Levi
------------------------------

#DecisionOptimization

I'm sorry, somehow I forgot to mention that I'm using c++, and if it helps, I'm using Eigen vars.

I think I unconsciously perceived cplex as having a built-in cpp in it :)

------------------------------
Zohar Levi
------------------------------

Original Message:
Sent: Wed August 17, 2022 04:20 AM
From: Vincent Beraudier
Subject: Solve a linear programming in canonical form

You may have a look at the docplex Python package. 

It provides some ways to create a model from a sparse matrix (numpy)
See http://docs.cplex.org.cn/mp/docplex.mp.advmodel.html?highlight=advmodel and the methods matrix_constraints.

You can also use a transformers from docplex: https://ibmdecisionoptimization.github.io/docplex-doc/2.21.207/mp/docplex.mp.sktrans.transformers.html?highlight=cplextransformer#docplex.mp.sktrans.transformers.CplexTransformerBase
and an example here: https://stackoverflow.com/questions/50714397/cplex-with-python-api-how-to-make-model-formulating-faster/50716526#50716526

------------------------------
Vincent Beraudier

Original Message:
Sent: Mon August 15, 2022 06:41 PM
From: Zohar Levi
Subject: Solve a linear programming in canonical form

I have a canonical problem:

min c'x
s.t. Ax <= b

It's a large problem using a sparse matrix. 
What's the easiest way to formulate it in cplex? Since it's already in canonical form, I was hoping for a low level function that can accept these vars instead of using the modeling interface.

------------------------------
Zohar Levi
------------------------------

#DecisionOptimization

We don't provide such construction in C++.
you will have to build the CPLEX model yourself. The code should look like this, I think:

SparseMatrix<double> mat = ...;
double[] d = ...;

IloModel model(env);
IloNumVarArray vars(env, mat.cols());

// to help build the constraint expression for each row.
IloNumVarArray selected_vars(env);
IloNumArray selected_coefs(env);

for (int k=0; k<mat.outerSize(); ++k){
  selected_vars.clear();
  selected_coefs.clear();
  for (SparseMatrix<double>::InnerIterator it(mat,k); it; ++it)
  {
    selected_coefs.add( it.value() );
    selected_vars.add( it.col() );
  }
  model.add( IloScalProd(selected_vars, selected_coefs) <= d[k] );
}
selected_vars.end();
selected_coefs.end();​

------------------------------
Vincent Beraudier
------------------------------

Original Message:
Sent: Wed August 17, 2022 06:43 AM
From: Zohar Levi
Subject: Solve a linear programming in canonical form

I'm sorry, somehow I forgot to mention that I'm using c++, and if it helps, I'm using Eigen vars.

I think I unconsciously perceived cplex as having a built-in cpp in it :)

------------------------------
Zohar Levi

Original Message:
Sent: Wed August 17, 2022 04:20 AM
From: Vincent Beraudier
Subject: Solve a linear programming in canonical form

You may have a look at the docplex Python package. 

It provides some ways to create a model from a sparse matrix (numpy)
See http://docs.cplex.org.cn/mp/docplex.mp.advmodel.html?highlight=advmodel and the methods matrix_constraints.

You can also use a transformers from docplex: https://ibmdecisionoptimization.github.io/docplex-doc/2.21.207/mp/docplex.mp.sktrans.transformers.html?highlight=cplextransformer#docplex.mp.sktrans.transformers.CplexTransformerBase
and an example here: https://stackoverflow.com/questions/50714397/cplex-with-python-api-how-to-make-model-formulating-faster/50716526#50716526

------------------------------
Vincent Beraudier

Original Message:
Sent: Mon August 15, 2022 06:41 PM
From: Zohar Levi
Subject: Solve a linear programming in canonical form

I have a canonical problem:

min c'x
s.t. Ax <= b

It's a large problem using a sparse matrix. 
What's the easiest way to formulate it in cplex? Since it's already in canonical form, I was hoping for a low level function that can accept these vars instead of using the modeling interface.

------------------------------
Zohar Levi
------------------------------

#DecisionOptimization

That should do, thanks.

------------------------------
Zohar Levi
------------------------------

Original Message:
Sent: Wed August 17, 2022 10:56 AM
From: Vincent Beraudier
Subject: Solve a linear programming in canonical form

We don't provide such construction in C++.
you will have to build the CPLEX model yourself. The code should look like this, I think:

SparseMatrix<double> mat = ...;double[] d = ...;IloModel model(env);IloNumVarArray vars(env, mat.cols());// to help build the constraint expression for each row.IloNumVarArray selected_vars(env);IloNumArray selected_coefs(env);for (int k=0; k<mat.outerSize(); ++k){  selected_vars.clear();  selected_coefs.clear();  for (SparseMatrix<double>::InnerIterator it(mat,k); it; ++it)  {    selected_coefs.add( it.value() );    selected_vars.add( it.col() );  }  model.add( IloScalProd(selected_vars, selected_coefs) <= d[k] );}selected_vars.end();selected_coefs.end();​

------------------------------
Vincent Beraudier

Original Message:
Sent: Wed August 17, 2022 06:43 AM
From: Zohar Levi
Subject: Solve a linear programming in canonical form

I'm sorry, somehow I forgot to mention that I'm using c++, and if it helps, I'm using Eigen vars.

I think I unconsciously perceived cplex as having a built-in cpp in it :)

------------------------------
Zohar Levi

Original Message:
Sent: Wed August 17, 2022 04:20 AM
From: Vincent Beraudier
Subject: Solve a linear programming in canonical form

You may have a look at the docplex Python package. 

It provides some ways to create a model from a sparse matrix (numpy)
See http://docs.cplex.org.cn/mp/docplex.mp.advmodel.html?highlight=advmodel and the methods matrix_constraints.

You can also use a transformers from docplex: https://ibmdecisionoptimization.github.io/docplex-doc/2.21.207/mp/docplex.mp.sktrans.transformers.html?highlight=cplextransformer#docplex.mp.sktrans.transformers.CplexTransformerBase
and an example here: https://stackoverflow.com/questions/50714397/cplex-with-python-api-how-to-make-model-formulating-faster/50716526#50716526

------------------------------
Vincent Beraudier

Original Message:
Sent: Mon August 15, 2022 06:41 PM
From: Zohar Levi
Subject: Solve a linear programming in canonical form

I have a canonical problem:

min c'x
s.t. Ax <= b

It's a large problem using a sparse matrix. 
What's the easiest way to formulate it in cplex? Since it's already in canonical form, I was hoping for a low level function that can accept these vars instead of using the modeling interface.

------------------------------
Zohar Levi
------------------------------

#DecisionOptimization