LANS Seminar

Seminar Title: A Review of “Faster Linear Algebra with Structured Random Matrices”

Speaker: Vishwas Rao, Computational Mathematician

Affiliation: Argonne National Laboratory, MCS Division

Date: Thursday, July 9, 2026

Time: 2:30 PM-3:30 PM (In-Person)

Location: Hybrid, Bldg. 240, Conference Room 4301

Zoom Link: https://argonne.zoomgov.com/j/1659346308?pwd=peKBZ2iAXFz4TTxuveQukvfIUfmqxb.1

Meeting ID: 165 934 6308

Passcode: 448219

Abstract: This is an expository talk reviewing a recent paper by Camaño, Epperly, Meyer, and Tropp; none of the results are my own. Randomized algorithms for the SVD, Nyström approximation, and least squares all rely on the same trick: replace A large matrix A with a small random sketch. In theory, the random test matrix is a dense Gaussian. In practice, people reach for much cheaper structured alternatives – sparce matrices, fast trigonometric transforms, Khatri-Rao products – and they work just as well. We’ll look at these cheaper alternatives through two lenses for what it means for a sketch to “work”: oblivious subspace embeddings (OSEs) and oblivious subspace injections (OSIs).

Bio: Vishwas Rao is a Computational Mathematician at the MCS division.  He has been at Argonne National Laboratory since 2017.

See all upcoming talks at https://www.anl.gov/mcs/lans-seminars