CODEWITHSUNDEEP
c++matrixeigeneigen3
GeneGi
GeneGi

Reputation: 426

eigen error: SelfAdjointView is only for squared matrices

I want to solve a sparse linear system like this:

SparseMatrix<double> A(m, n);
VectorXd b(m);
ConjugateGradient<SparseMatrix<double>, Upper> solver;
solver.compute(A);
VectorXd X = solver.solve(b);

but I got this error for running this code:

Assertion failed: (rows()==cols() && "SelfAdjointView is only for squared matrices"), function SparseSelfAdjointView

why I got this problem and how to solve it?

I write a small example to reproduce this error:

#include "lib/Eigen/Sparse"

using namespace Eigen;

int main()
{
    SparseMatrix<double> A(2, 3);

    A.coeffRef(0, 0) = 1;
    A.coeffRef(0, 1) = 1;
    A.coeffRef(0, 2) = 1;
    A.coeffRef(1, 0) = 1;
    A.coeffRef(1, 1) = 1;
    A.coeffRef(1, 2) = 1;

    VectorXd b(2);
    b << 3, 3;

    ConjugateGradient<SparseMatrix<double>, Upper> solver;
    solver.compute(A);
    VectorXd X = solver.solve(b);

    return 0;
}

Upvotes: 0

Views: 160

Answers (1)

chtz
chtz

Reputation: 18807

The ConjugateGradient algorithm only works with self-adjoint matrices. For rectangular matrices use LeastSquaresConjugateGradient instead.

Upvotes: 2

Related Questions