MatrixXf m(3,2);

m <<

1.0, 2.0,

3.0, 4.0,

5.0, 6.0

;

JacobiSVD<MatrixXf> svd(m, ComputeFullU | ComputeFullV);

auto e = svd.singularValues();

// TODO. e needs to be 3x2, what is the nicest way? Almost there, but this changes value positions.

//MatrixXf e(svd.singularValues().asDiagonal());

//e.resize(m.rows(), m.cols());

cout << "E" << endl << e << endl << endl;

auto u = svd.matrixU();

cout << "U" << endl << u << endl << endl;

cout << "UU'" << endl << u.adjoint() * u << endl << endl;

auto v = svd.matrixV();

cout << "V" << endl << v << endl << endl;

cout << "VV'" << endl << v * v.adjoint() << endl << endl;

//cout << "UEV" << endl << u * e * v << endl << endl;

cout << "least squares" << endl << svd.solve(Vector3f(1, 0, 0)) << endl << endl;