Closed dylan-copeland closed 4 years ago
@dylan-copeland This is awesome! Is this ready for review?
@dylan-copeland This is awesome! Is this ready for review?
@chldkdtn Not ready yet. There will also be a small PR for Laghos to use QDEIM. We should start using PR labels (e.g. "Work-in-progress", "Ready-for-review").
That is a good idea. I have labeled this as "WIP"
@chldkdtn @kevinhkhuynh reminder that this is the only QDEIM PR waiting for your review.
QDEIM is implemented, as an alternative to DEIM and GNAT, with oversampling done by GappyPOD+E (see https://arxiv.org/abs/1808.10473). The first samples (numbering up to the basis dimension) for QDEIM are computed using parallel QR with column pivoting for the transposed basis matrix, via
pdgeqpf
in scalapack.The only validation of this code is its application in Laghos, which has some good results but does not truly validate the QR or QDEIM implementations.