You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This gives you access to the ULV factors for an HSSMatrixMPI: 94f8f3a
I'll need to think about how to compute the determinant.
In the original ULV factorization, U and V were orthogonal, then you can get the determinant from the L factor.
But our implementation of ULV here is based on this work: https://epubs.siam.org/doi/abs/10.1137/12087116X
And the HSS matrix has special structure which is exploited in the ULV factorization.
@pghysels . In my specific application the matrix is a covariance matrix, so A = A^T. This should simplify the determinant calculation, as that only depends on the determinants of the leaves and the symmetric matrices formed by X X^T of the off-diagonals at each node. I have not understood the algorithm not implementation enough to figure out how to do that from the factorization though.
The XIa reference works as an example a positive definite matrix, so the solution does seem to come from getting the diagonals of L_ii and recursively getting the determinant of the script(U) term in Eq. 3.12.
I am working with very large matrices, for which there is currently no HSSMatrixMPI<>.ULV().
I would like the log determinant of an HSSMatrix / HSSMatrixMPI.
The text was updated successfully, but these errors were encountered: