User:Christopher G. Baker: Difference between revisions
imported>Christopher G. Baker No edit summary |
imported>Christopher G. Baker No edit summary |
||
| Line 6: | Line 6: | ||
My dissertation concerns optimization on Riemannian manifolds. More specifically, I am interested in the class of retraction-based optimization methods, particularly the Riemannian trust-region methods. Feel free to visit [http://www.scs.fsu.edu/~cbaker my home page] for more information on this topic, as well as links to publications. | My dissertation concerns optimization on Riemannian manifolds. More specifically, I am interested in the class of retraction-based optimization methods, particularly the Riemannian trust-region methods. Feel free to visit [http://www.scs.fsu.edu/~cbaker my home page] for more information on this topic, as well as links to publications. | ||
[[Category:CZ Authors]] | |||
Revision as of 11:24, 14 February 2007
My name is Christopher G. Baker. I am a Ph.D. candidate in Computer Science at Florida State University. I am currently participating in an internship at Sandia National Laboratories in Albuquerque, NM, while I complete my dissertation.
My work at Sandia is on high-performance, robust parallel algorithms in the Trilinos project. Trilinos is a collection of large-scale solvers: linear systems, eigenvalue problems, non-linear optimization. My principal work is on Anasazi, the block eigensolvers package.
My master's thesis was entitled "A Block Incremental Algorithm For Computing Dominant Singular Subspaces." In this work, I described and analyzed a family of methods for incrementally computing low-rank approximations of a matrix, based on the truncated SVD. This should be contrasted with the common technique of computing the full SVD of a matrix (either incrementally or as a batch) and truncating the unwanted part of the factorization.
My dissertation concerns optimization on Riemannian manifolds. More specifically, I am interested in the class of retraction-based optimization methods, particularly the Riemannian trust-region methods. Feel free to visit my home page for more information on this topic, as well as links to publications.