2–5 Jul 2024
Osijek
Europe/Zagreb timezone

Randomized algorithms with random Khatri-Rao product matrices

Not scheduled
20m
Osijek

Osijek

School of Applied Mathematics and Informatics, J. J. Strossmayer University of Osijek, Trg Ljudevita Gaja 6, Osijek Faculty of Economics, J. J. Strossmayer University of Osijek , Trg Ljudevita Gaja 7, Osijek
Invited lecture NA: Numerical Analysis and Scientific Computing

Speaker

Zvonimir Bujanović (University of Zagreb, Faculty of Science, Department of Mathematics)

Description

Randomized algorithms have gained significant attention in numerical linear algebra during the last decade. In particular, randomized sketching is used as a simple but effective technique in which a random matrix acts as a dimension reduction map: a problem that features a potentially large input matrix $A \in \mathbb{R}^{n \times n}$ is reduced to a smaller one by replacing $A$ with $A \Omega$, where $\Omega \in \mathbb{R}^{n \times \ell}$ is a random matrix with $\ell \ll n$. This has been used with great success in, e.g., randomized SVD and subspace projection methods for large-scale eigenvalue problems, such as FEAST.

Algorithms based on sketching typically draw random matrices $\Omega$ from standard distributions, such as Gaussian. However, in certain applications it may be advantageous to run computations with $\Omega$ that is compatible with the underlying structure of the problem. In this talk we discuss algorithms that use random Khatri-Rao product matrices: each column of $\Omega$ is generated as the Kronecker product of two Gaussian random vectors. This will allow for faster operations when the matrix $A$ is represented as a short sum of Kronecker products, which arises frequently, e.g., from the discretization of PDEs on tensor product domains. We focus on applications in large-scale eigenvalue computation, and provide theoretical and numerical evidence that the use of random Khatri-Rao product matrices $\Omega$ instead of unstructured Gaussian random matrices leads to good estimates.

This is joint work with Luka Grubišić, Daniel Kressner and Hei Yin Lam.

Primary author

Zvonimir Bujanović (University of Zagreb, Faculty of Science, Department of Mathematics)

Presentation materials

There are no materials yet.