On 10/15/19 5:17 PM, Maien Hamed wrote: > > But this is not reliable, as I've run into situations where the returned > eigenvalues are fewer than `dim` (I'm not sure how > `compute_eigenvalues_symmetric` treats repeated eigenvalues). I've looked at > using `compute_svd`, but it seems to me that I could only access the > `singular_value`s that way, but not the `svd_u` and `svd_vt` matrices.
For matrices of size 2x2 and 3x3, you could of course compute the eigenvalues more easily than by using LAPACK -- in fact, I think we have code for that somewhere (maybe in the FullMatrix class? -- I don't recall). But the point of course remains that the log of a matrix is really only well defined if the eigenvalues are in fact bounded away from zero. If they're too small, you run into problems with the log being a large negative number. It may also be possible to come up with an entirely different way to compute the log than via the eigenvalue/vector decomposition. There is a rather well known paper that is, I think, called "11 ways to compute the matrix exponential" or something of this sort. There might be an equal number of ways computing the logarithm, but I really wouldn't know. Best W. -- ------------------------------------------------------------------------ Wolfgang Bangerth email: bange...@colostate.edu www: http://www.math.colostate.edu/~bangerth/ -- The deal.II project is located at http://www.dealii.org/ For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en --- You received this message because you are subscribed to the Google Groups "deal.II User Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to dealii+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/dealii/459036b1-13c0-1789-8db6-ab564ec0472c%40colostate.edu.