DATE:

Friday, April 30, 2021

TIME:

3:30 PM

LOCATION:

Virtual Zoom Conference

SPEAKER:

Calvin Johnson, Physics, San Diego State University

ABSTRACT:

A common numerical problem is to find eigenvalues and eigenvectors of symmetric or Hermitian matrices–especially extremal eigenvalues. An example is in the matrix representation of the many-body Schrodinger equation. While many applied math journal articles on eigensolvers may discuss examples of matrices of dimensions of a few million, in our applications my group has to handle matrices with dimensions in hundreds of millions and even billions. How can one possibly tackle such huge problems? How do you even store the data? I discuss the techniques we use, starting with the Lanczos algorithm, a prominent example of Arnoldi methods, and how we efficiently carry these out on laptops, workstations, and supercomputers. The full story provides a classic example of the interweaving of science and computation.

HOST:

Samuel Shen

VIDEO: