[FRIAM] Looking for an algorithm

[Jon Zingale]

Thanks to everyone who helped pitch in on this. I was happy to finally
track down a pdf of Van Loan & Pitsianis [1993]. As stated in this other
article on "Automated Kronecker Product Approximation"[KoPA]:

"This was introduced in the matrix computation literature as the nearest
Kronecker product problem in Van Loan and Pitsianis, who demonstrated its
equivalence to the best rank-one approximation and therefore also to the
SVD, after a proper rearrangement of the matrix entries."

As well as:

"Finding a low-rank approximation of a given matrix is closely related to
the singular value decomposition, and the connection was revealed as early
as Eckart and Yount (1936)]"

It is surprising that it took until the 90s for the problem to bite someone
and for explicit algorithms to emerge. Also, I am amazed that the body of
the literature seems to come from the data sciences, as I arrived at the
problem from studying discrete dynamical systems. I am hoping to write up a
paper on the connection between them and NKP soon. Would anyone know where
I can find a python, C++, or Haskell implementation easily?

Thanks again for the help.

[1993]
https://www.cs.cornell.edu/cv/ResearchPDF/Approximation%20with%20Kronecker%20Products,%20from%20Linear%20Algebra%20for%20Large%20Scale%20and%20Real-%20Time%20Applications.pdf

[KoPA] https://arxiv.org/pdf/1912.02392.pdf
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20211104/5fdc9ce4/attachment.html>