Giuseppe
Ancona
Strasbourg
Ramified periods and field of definition
Search by speaker
Filter institution
Filter content
20. October
Tomorrow
Akif
Çördük
NVIDIA
GPU-Accelerated Optimization: Primal Heuristics and Multi-GPU PDLP
Abstract.
In this talk, we’ll look at how GPUs can help find feasible solutions to mixed-integer programs and solve large linear programs. We’ll cover three primal heuristics—a feasibility pump using PDLP, fix-and-propagate, and feasibility jump—and discuss the practical challenges of making them run efficiently on GPUs. We’ll then turn to multi-GPU PDLP and how to partition the problem to balance work across devices while keeping communication costs low.
21. October
Wed W42
Kolyan
Ray
Imperial College London
Bayesian regression models with dense confounding variables
Abstract.
Inferring causal relationships from observational data can be invalidated by the existence of confounding variables. We consider nonparametric Bayesian inference in regression models with additive errors under a dense confounding assumption, i.e. that every confounding variable affects many covariates. We show that by using suitable priors, which are equivalent to marginalizing out the confounders in a certain sense, such confounding can be corrected for. Our approach is illustrated via Gaussian process and spike and slab priors, for which we provide accompanying theoretical guarantees via posterior convergence rates. This is joint work with Luke Travis
Lenny
Fukshansky
Claremont McKenna College
Geometric constructions for sparse integer signal recovery
Ilia
Itenberg
Institut de Mathématiques de Jussieu
Topology of real tropical hypersurfaces
Abstract.
Real algebraic varieties seem to be quite distant from combinatorial geometry. However, it is possible to construct real algebraic hypersurfaces in projective spaces (and, more generally, in toric varieties) in a completely combinatorial fashion: one can patchwork them from pieces which essentially are hyperplanes. This procedure, called the combinatorial patchworking, is a particular case of the Viro method of construction of real algebraic varieties with prescribed topology. The combinatorial patchworking is one of the sources of tropical geometry, and the results of this construction can be seen as real tropical hypersurfaces. We will discuss the combinatorial patchworking and several results concerning the topology of real tropical hypersurfaces.
22. October
Thu W42
Clara
Thierbach
HU Berlin
A variational characterization of a Gibbs measure related to the Aviles-Giga functional
Abstract
Bashar
Hamade
Formalizing the Edge Theorem for Robust Control Stability in Lean4
Abstract.
In control theory, the stability of a system is often analyzed through the roots, or poles, of its transfer-function polynomial. When polynomial coefficients vary over prescribed ranges, robust stability can be studied using results such as Kharitonov’s Theorem for continuous-time systems and the Edge Theorem for discrete-time systems. In this project, we formalize the Edge Theorem and the supporting results in Lean 4. The development uses polyhedral geometry and root-space properties to show how roots associated with a polytope of polynomials relate to those of its exposed faces and edges, and how the relevant boundary results follow.
23. October
Fri W42
Stefanie
Petermichl
U Würzburg
26. October
Mon W43
Raphael
Kuess
HU Berlin
27. October
Tue W43
Alexander
Bobenko
TU Berlin
Peter
Buergisser
Berlin
28. October
Wed W43
Olivier
de Gaay Fortman
Imperial College
Nicolas
Verzelen
INRAE, Université Montpellier
Statistical–Computational Trade-offs: Low-Degree Polynomials and Community Recovery
Abstract.
In many high-dimensional problems, such as planted clique, sparse PCA, clustering and community detection, the signal strength needed by the best known polynomial-time algorithms is much larger than what is information-theoretically necessary. Is this gap intrinsic? Since NP-hardness says nothing about random instances, evidence of hardness must come from other sources: average-case reductions, or lower bounds for broad classes of algorithms. In the first part, I will introduce the low-degree polynomial framework, starting from testing problems in additive models. In the Gaussian additive model, the χ² divergence and the low-degree advantage admit simple expressions in terms of the overlap between two independent copies of the signal. Comparing them shows how truncating at a low degree discards the rare large overlaps that only exhaustive search can exploit. In the second part, based on joint work with Alexandra Carpentier and Christophe Giraud, I will consider community recovery in the stochastic block model with more than √n communities. We prove that low-degree polynomials fail below a threshold conjectured by Chin, Mossel, Sohn and Wein, for any graph density, and that this threshold lies below the Kesten–Stigum threshold. We also show that counting suitable motifs succeeds above it in moderately sparse regimes. These motifs are dictated by the lower-bound analysis: cliques, paths, and blow-ups of cycles.
Lars
Schroeder
University of Twente
The Geometry of Hierarchical Navigation
Abstract.
Many applications use multi-layered hierarchical data structures for fast approximate nearest-neighbor search in high-dimensional vector spaces. However, the geometric conditions that ensure accurate and efficient greedy navigation remain poorly understood. In this talk, we study the efficiency of greedy navigation on a hierarchy of proximity graphs constructed from n data points on the $d$-dimensional torus. We identify a deterministic coverage condition under which, given any query $q$, greedy search returns a point within $(1+\epsilon)$-factor of the distance to the closest point. This coverage property holds with high probability when the data is distributed as a homogeneous Poisson process, a Hermitian determinantal process, or a bounded-density Cox process, as long as $d=o(\log n/\log\log n)$. Additionally, we obtain logarithmic expected hop count for the number of greedy hops in fixed dimension. Joint work with Shankar Bhamidi, Souvik Dhara and Clara Stegehuis.
Eugenia
Saorín Gómez
Orthogonal projections, linearization and sausages: a geometric approach to Cheeger sets
Abstract
29. October
Thu W43
Cesar
Ceballos
Carlo
Sgarra
Università degli Studi di Bari Aldo Moro
Semi-static variance-optimal hedging with self-exciting jumps
Abstract.
The aim of this talk is to investigate a quadratic, i.e., variance-optimal, semi-static hedging problem in an incomplete market model where the underlying log-asset price is driven by a diffusion process with stochastic volatility and a self-exciting jump process of Hawkes type. More precisely, we aim at hedging a claim at time T > 0 by using a portfolio of available contingent claims, so to minimize the variance of the residual hedging error at time T. In order to improve the replication of the claim, we look for a hybrid hedging strategy of semi-static type, in which some assets are continuously rebalanced (the dynamic hedging component) and for some other assets a buy-and-hold strategy (the static component) is performed. We discuss in detail a specific example in which the approach proposed is applied, i.e., a variance swap hedged by means of European options, and we provide a numerical illustration of the results obtained. (In cooperation with G. Callegaro, P. Di Tella and B. Ongarato, to appear on Mathematics of Operations Reasearch)
Philipp
Schmocker
ETH
Smooth structures on spaces of probability measures
Abstract.
We introduce a categorical framework for higher-order probability theory based on quasi-M-spaces and Frölicher M-spaces. Besides Giry's and Forré's probability monads, we introduce our own hybrid pushforward probability monad generated by randomizations of a fixed reference measure. These constructions are canonically transported to saturated Frölicher M-spaces, yielding, in the measurable setting, Giry and quasi-probability monads on the corresponding Frölicher spaces. In the smooth setting, the resulting spaces of probability measures carry a natural smooth structure, which can be related to the continuity equation, Otto calculus and Lions' derivative. This is joint work with Philipp Harms, Peter W. Michor and Josef Teichmann.
Christoph
Czichowsky
LSE
03. November
Tue W44
Nitin Kumar
Chidambaram
UNED Madrid
Refining Witten-Kontsevich
Abstract.
The moduli space of metric Möbius graphs, which are non-orientable ribbon graphs, has one component homeomorphic to the moduli space of Riemann surfaces and another component homeomorphic to the moduli space of Klein surfaces. I'll discuss a lattice point count on this moduli space, weighted by a polynomial in b known as the measure of non-orientability. This "refined" lattice point count satisfies a refined version of Norbury's recursion for the count of lattice points on the moduli space of curves. Consequently, we obtain a recursion for the volumes of these moduli spaces that reduces to the Witten-Kontsevich recursion when b=0.
Charles
Vial
Bielefeld
04. November
Wed W44
Mark
Podolski
Universität Luxemburg
Martin
Kružík
Czech Academy of Sciences Prague, CZ
The effects of pressure loads in the dimension reduction of elasticity models
Abstract
05. November
Thu W44
Alex
Lassota
Patrick
Heimbach
Oden Institute for Computational Engineering and Sciences, University of Texas at Austin
Differentiable programming for hybrid Earth system modeling
Abstract.
Despite the revolution in autonomous in-situ and satellite remote sensing, much of the ocean remains poorly sampled observationally in terms of its spatio-temporal evolution. Inverse methods seek to combine the incomplete knowledge reservoirs from sparse observations and numerical models that require the provision of uncertain inputs to produce optimal state and parameter estimates of the system considered. In its various renderings, methods of optimal estimation and control or data assimilation are being used to learn uncertain parameters (calibration problem) in models or to produce optimal initial states for prediction (extrapolation problem). State estimates serve as optimal reconstructions to improve system understanding (interpolation problem). Rigorously applied, these methods also enable characterization of uncertainty that arise from different sources (model, observations, forcings). Scientific machine learning methods rely on similar mathematical concepts that underpin optimal estimation and may be successful when the system is heavily sampled. In this presentation I explore some of the concepts in the context of ocean modeling, in particular the concept of differentiable programming. I provide a subjective perspective on opportunities and challenges with a view toward hybrid data assimilation and machine learning to enable online or full-model learning. These ideas are being implemented in the context of general-purpose automatic differentiation applied to a new ocean model written in the Julia programming language.
06. November
Fri W44
10. November
Tue W45
Davide
Scazzuso
HU Berlin
A novel open/closed duality on toric Calabi--Yau threefolds
Abstract.
The study of A-model topological string theory on toric Calabi--Yau threefolds X occupies a distinguished position at the crossroads of physics and mathematics, as it provides a tractable framework for string theory in which quantum string amplitudes can be defined rigorously and computed exactly. In this talk, I will discuss the open-closed A-model on X in the presence of Aganagic--Vafa special Lagrangian A-branes, as well as the closed A-model on X in the presence of Ramond--Ramond fluxes induced by toric B-branes. I will establish an open-closed duality theorem that explicitly relates the amplitudes arising in these two setups, thereby demonstrating that the two A-model configurations provide dual descriptions of the same underlying physical and enumerative problem. If time permits, I will further comment on the conjectural interpretation of the duality in terms of the geometric engineering of surface and line defects in four-dimensional N=2 gauge theory.
Gregorio
Baldi
Jussieau
Dominic
Bunnett
Berlin
11. November
Wed W45
Simon
Weißmann
Universität Mannheim
12. November
Thu W45
Rohit
Pochampalli
HU Berlin
Federica
Masiero
University of Milano-Bicocca
17. November
Tue W46
Ivan
Dneprov
University of Mons
Óscar
Fenández Ayala
Braunschweig
18. November
Wed W46
Botond
Szabo
Bocconi University
Vinicio
Rodrigues Moreira
UFMG Brazil
Vinicio Rodrigues
Moreira
UFMG Brazil
Gianna
Götzmann
U Augsburg
Perturbed minimizing movements of time-dependent functionals on metric spaces
Abstract
19. November
Thu W46
Nathanaël
Berestycki
Universität Wien
Near-critical dimers and sine-Gordon model
Abstract.
We consider the dimer model with Temperleyan boundary conditions on the square lattice (or more generally isoradial superpositions). When the edge weights are critical it is well known since the work of Kenyon that in the scaling limit, the associated height function is conformally invariant and is described by the Gaussian free field (aka massless free boson). We consider here a massive perturbation of this model, where the edge weights deviate slightly from their critical values. I will discuss a recent result, obtained jointly with Scott Mason and Lucas Rey, in which we showed that the associated height function is described in the scaling limit by the sine-Gordon model at its free fermion point. In parallel, geometric interfaces naturally associated via Temperley's bijection were previously shown (in joint work with Levi Haunschmid-Sibitz) to converge to a massive variant of SLE (Schramm--Loewner evolution) introduced by Makarov and Smirnov. We will strive to present these ideas in a non-technical manner; in particular no probability background should be needed.
24. November
Tue W47
Javier
Fresán
Jussieau
25. November
Wed W47
Christoph
von Tycowicz
Zuse Institute Berlin (ZIB)
Symmetry-preserving geodesic regression on Lie Groups for longitudinal medical imaging
Abstract.
Many medical imaging tasks require statistical modeling of continuous transformations, including longitudinal anatomical shape change and articulated skeletal motion. These transformations naturally live on Lie groups, where meaningful statistical analysis should respect group symmetries to remain invariant to arbitrary coordinate choices and reference frames. In this talk, I will present a geodesic regression framework on Lie groups for longitudinal imaging data. Common approaches rely on Riemannian metrics, but many Lie groups do not admit a metric fully compatible with the group structure. This mismatch breaks symmetry and leads to unstable regression estimates. We therefore introduce a non-metric, bi-invariant estimator that is equivariant under both left and right group actions. We evaluate the method on synthetic data and on an open-access clinical dataset of longitudinal knee joint configurations acquired for osteoarthritis research. The proposed approach yields stable trajectories and reproducible statistical conclusions, while state-of-the-art Riemannian methods exhibit sensitivity and instability. These results highlight the practical advantages of symmetry-preserving statistical modeling in longitudinal medical imaging studies.
Ulrik
Thinggaard Hansen
Karlsruher Institut für Technologie
Yan
Alves Radtke
Fast Counting of Cyclic Polytope Triangulations
26. November
Thu W47
Purba
Das
King's College London
Matthieu
Meunier
Oxford
27. November
Fri W47
Mario
Kummer
TU Dresden
01. December
Tue W48
Amanda
Hirschi
MPIM Bonn
Luca Maria
Di Gravina
Halle
02. December
Wed W48
Hugo
Cui
Laboratory of Orsay
Hugo
Cue
Orsay
04. December
Fri W48
08. December
Tue W49
Lukas
Müller
LMU München
09. December
Wed W49
Yuta
Koike
Tokyo University
Kathrin
Hellmuth
Caltech, USA
10. December
Thu W49
Giulia
Liveri
LSE
Giuseppina
Guatteri
Polytechnic University of Milan
11. December
Fri W49
Stefanie
Petermichl
U Würzburg
15. December
Tue W50
Marita
Thomas
FU Berlin
16. December
Wed W50
Eardi
Lila
University of Washington
Geometric deep operator learning for inverse problems with functional data
25. December
Fri W51
Fri, 25. Dec
Christmas Day
26. December
Sat W51
Sat, 26. Dec
St. Stephen's Day
01. January
Fri W00
Fri, 01. Jan
New Year's Day
06. January
Wed W01
Claudia
Strauch
Universiät Heidelberg
08. January
Fri W01
Bernhard
Schmitzer
Göttingen
12. January
Tue W02
13. January
Wed W02
Shahin
Tavakoli
University of Geneva
Matthias
Hieber
TU Darmstadt
22. January
Fri W03
László
Végh
Bonn
26. January
Tue W04
Amanda
Hirschi
MPIM Bonn
28. January
Thu W04
Katia
Colaneri
University of Rome Tor Vergata
10. February
Wed W06
Alexander
Kreiß
Universität Hildesheim
17. February
Wed W07
Shahin
Tavakoli
University of Geneva
08. March
Mon W10
Mon, 08. Mar
International Women's Day (Regional Holiday)
26. March
Fri W12
Fri, 26. Mar
Good Friday