Feature/spectra correlations #306
Conversation
include the spectra implementation in the NT minPosLinearEigenvalue_EigenSymSolver
this function checks if a matrix is indeed a correlation matrix
| @@ -168,8 +168,9 @@ class EigenvaluesProblems<NT, Eigen::Matrix<NT,Eigen::Dynamic,Eigen::Dynamic>, E | |||
| Spectra::DenseCholesky<NT> Bop(-A); | |||
|
|
|||
| // Construct generalized eigen solver object, requesting the largest three generalized eigenvalues | |||
| int ncv = std::min(std::max(10, matrixDim/20), matrixDim); | |||
There was a problem hiding this comment.
Could you please comment on where those numbers come from? 10, 20, 15 ?
There was a problem hiding this comment.
This is an empirical value I gave to @atrayees. We left as a task after the R interface to tune this implementation (not only ncv value).
There was a problem hiding this comment.
Thanks @TolisChal for the explanations. @atrayees could you please add a comment with this explanation and mention that as a TODO?
| } | ||
| } | ||
|
|
||
| //check if the matrix is positive definite |
There was a problem hiding this comment.
Why not using the function isPositiveSemidefinite. Note: that function is not using tolerance.
There was a problem hiding this comment.
I have updated the function and made use of the isPositiveSemidefinite function
|
|
||
| if(eigen_solver.info() != Eigen::Success) | ||
| { | ||
| return false; |
| int valid_points = 0; | ||
| for(const auto& points : randPoints){ | ||
| if(is_correlation_matrix(points.mat)){ | ||
| valid_points++; |
| @@ -168,8 +168,11 @@ class EigenvaluesProblems<NT, Eigen::Matrix<NT,Eigen::Dynamic,Eigen::Dynamic>, E | |||
| Spectra::DenseCholesky<NT> Bop(-A); | |||
|
|
|||
| // Construct generalized eigen solver object, requesting the largest three generalized eigenvalues | |||
There was a problem hiding this comment.
"... computing the minimum positive eigenvalue by computing the largest eigenvalue of the inverse Generalized Eigenvalue Problem."
There was a problem hiding this comment.
i updated the comment.
I made the following changes-
volesti/matrix_operations/EigenvaluesProblems.hfrom 15 to a more useful valueminPosLinearEigenvalue_EigenSymSolverfunction