Hi,

For dense matrix, one could use code like:

Code: Select all

MatrixXd m = MatrixXd::Random(1000,1000);

Although Eigen could define a large sparse matrix like:

Code: Select all

Eigen::SparseMatrix<double> M(1000, 1000);

How could I fill it with Random values?

Thank you.

There is no convenient methods because the structure of the sparse matrix is a very important aspect, and it usually does not make much sense to initialize a sparse matrix with a random structure.

ggael wrote:There is no convenient methods because the structure of the sparse matrix is a very important aspect, and it usually does not make much sense to initialize a sparse matrix with a random structure.

By using Armadilo+Arpack, one could construct large Sparse Matrix with Random values like:
sp_cx_mat A (sprandu<sp_mat>(1000, 1000, 0.1), sprandu<sp_mat>(1000, 1000, 0.1));

It is convenient for me to testing the accuracy and efficiency of an algorithm.
Anyway, thank you all the same.

we have a similar helper in test/sparse.h that you can adapt to remove the construction of the dense reference matrix.

ggael wrote:we have a similar helper in test/sparse.h that you can adapt to remove the construction of the dense reference matrix.

Hi, Gael

Sorry for bothering you again.
I wrote a fragment of simple code to try, but it gives out fatal error C1083: Cannot open include file: 'split_test_helper.h': No such file or directory
Yet, I cannot find such file in the whole Eigen folder...

Code: Select all

#include <Eigen/Core>
#include <iostream>
#include <test/sparse.h>
using namespace Eigen;
using namespace std;
int main()
{
   MatrixXd A = MatrixXd::Random(1000, 1000);
   SparseMatrix<double> M(1000, 1000);
   initSparse(0.1, A, M);
   cout << M << endl;
   system("pause");
   return 0;
}

indeed, you cannot just directly include this file. Instead, you should rather copy/paste the related function and adjust it to your need.

For the record, split_test_helper.h is procedurally generated by cmake when configuring the unit test.