Hey everyone. As part of my PhD work I needed an implementation of the Band Lanczos Algorithm, so I implemented it from the "Templates for the Solution of Algebraic Eigenvalue Problems" book (
link) using eigen, but its not really a full blown template class. I figured it might be worth expanding and adding to the Unsupported set of libraries. I have a couple of questions about contributing, since it will still be a bit of work to make as generic as the rest of the decomposition classes.
1) Does this already exist somewhere?
2) How does Eigen deal with Hermitian/non-Hermitian naming, since usually a very similar algorithm exists for both. Should I make a separate class or just add a flag for whether the matrix is hermitian?
3) In these sorts of problems its sometimes possible that the Matrix being decomposed can't actually be formulated, so a function needs to passed that takes a vector and returns a vector. Are there any classes that already do this so I can make sure my API looks the same?
4) Which branch should I check out to add stuff to?
5) Can I just put the tests in unsupported/test?
Thanks
-Micah
--
Micah Chambers, M.S.
Ahmanson-Lovelace Brain Mapping Center
University of California Los Angeles
635 Charles E. Young Drive SW, Suite 225
Los Angeles, CA 90095-7334
310-206-2101 (Office)
310-206-5518 (Fax)
703-307-7692 (Cell)
email: micahcc@xxxxxxxx