Research Talks at the Zuse Institute Berlin
Welcome to the ZIB's Research Seminars and Colloquia page. Explore our schedule of research talks in mathematics and computer science. We currently host the following series at our institute:
- The Zuse Research Colloquium, a series of talks by high profile invited speakers from academia and industry.
- The Zuse Research Seminar, a series of talks by researchers from the Zuse Institute Berlin and select external speakers from the Berlin math community.
- The IOL Seminar and Lecture Series, the research seminar of the IOL research lab of Sebastian Pokutta at ZIB and TU Berlin.
For questions relating to an individual seminar or colloquium, please contact the organizers of that series. For question relating to this homepage, please contact Christoph Spiegel. The ZIB also hosts an overview of all mathematical research seminars happening in and around Berlin at seminars.zib.de.
Upcoming Talks
The Frank-Wolfe method is a classical first-order algorithm for constrained optimization that avoids projections by relying on linear minimization oracles. While extensively studied over polytopes, its behavior over strongly convex domains remains less understood, especially near the boundary. In this talk, we will explore the challenges of optimizing over such sets, with a particular focus on the difficulty of finding optima located on the boundary. Using the simple example of projections onto the L2-ball, we will provide geometric insights into convergence behavior of Frank-Wolfe. This example illustrates fundamental limitations of the method in boundary regimes and motivates further analysis of step size strategies and alternative update directions.
t.b.a.
t.b.a.
t.b.a.
Riemannian optimization refers to the optimization of functions defined over Riemannian manifolds. Such problems arise when the constraints of Euclidean optimization problems can be viewed as Riemannian manifolds, such as the symmetric positive-definite cone, the sphere, or the set of orthogonal linear layers for a neural network. This Riemannian formulation enables us to leverage the geometric structure of such problems by viewing them as unconstrained problems on a manifold. The convergence rates of Riemannian optimization algorithms often rely on geometric quantities depending on the sectional curvature and the distance between iterates and an optimizer. Numerous previous works bound the latter only by assumption, resulting in incomplete analysis and unquantified rates. In this talk, I will discuss how to remove this limitation for multiple algorithms and as a result quantify their rates of convergence.