Dates: Wednesday, April 16, 2025 - July 09, 2025
Organizers: Barbara Verfürth, Wolfgang Lück and Illia Karabash
Venue: Lipschitzsaal, Mathezentrum, Endenicher Allee 60, 53115 Bonn
Date
Hausdorff Tea
Hausdorff Colloquium
Graduate Colloquium
16.04.2025
15:00
15:15
Dominique Maldague (MIT, USA)
"An intersection of CS and harmonic analysis: approximating matrix p to q norms."
23.04.2025
15:00
15:15
Lars Becker (University of Bonn)
"Quantum signal processing and the nonlinear Fourier transform"
29.04.2025
Additional date, as an exception on a Tuesday
15:00
15:15
Hong Wang (New York University, USA)
TBA
30.04.2025
15:00
15:15
Sun Woo Park (MPIM)
"Graph neural networks and covering spaces"
21.05.2025
15:00
15:15
TBA
28.05.2025
15:00
15:15
László Székelyhidi (Max Planck Institute for Mathematics in the Sciences, Leipzig, Germany)
TBA
04.06.2025
15:00
15:15
Elliot Kaplan (MPIM)
TBA
25.06.2025
15:00
15:15
Habib Ammari (ETH, Zurich, Swiss)
TBA
02.07.2025
15:00
15:15
TBA
09.07.2025
15:00
15:15
Andreas Thom (TUD, Germany)
TBA
Dominique Maldague (MIT, USA): "An intersection of CS and harmonic analysis: approximating matrix p to q norms"
In Fourier restriction theory, we bound exponential sums whose frequencies lie in sets with special properties, e.g. random sets or curved sets. Bourgain, Demeter, and Guth developed decoupling inequalities to show that such functions enjoy square root cancellation behavior. This theory lies in the larger context of bounding matrix l^p to l^q norms, which is well-studied in the CS literature. We will discuss a new polynomial time algorithm inspired by Fourier restriction theory of myself, Guth, and Urschel which reduces the multiplicative error of computing matrix 2 to q norms from roughly n^{1/q} to n^{1/2q}, where n is the size of the matrix.
Lars Becker (University of Bonn): "Quantum signal processing and the nonlinear Fourier transform"
We will give a short overview of two topics and their connection. The first is so-called quantum signal processing, a framework for designing quantum algorithms. The second one is the nonlinear Fourier transform, a transformation that diagonalizes certain integrable nonlinear partial differential equations.
Sun Woo Park (MPIM): "Graph neural networks and covering spaces"
I would like to give a brief overview of some deep learning techniques, their applications, and their limitations. We will focus particularly on how covering spaces are relevant to understanding limitations of conventional neural networks in determining isomorphism classes of graphs (or in particular graph neural networks). A number of works presented in this talk are based on joint collaborations with Yun Young Choi, U Jin Choi, Dosang Joe, Minho Lee, Seunghwan Lee, Joohwan Ko, and Youngho Woo. My hope is to make the talk as accessible as possible, even for those who do not have prior knowledge in deep learning techniques.