Abstract: This workshop will bring together researchers working on the topology of algebraic varieties from several different points of view. Topics include algebraic cycles and motives, birational geometry, generalized cohomology theories, and classifying spaces and group cohomology.
Speaker: Bunyamin Sari (University of North Texas) - https://sites.cos.unt.edu/~bunyamin/ When: Mon, November 16, 2026 - 2:00pm Where: Phys 1219
View Abstract
Abstract: The Banach space $c_0$ is universal for separable metric spaces under bi-Lipschitz embeddings. However, $c_0$ admits neither a linear isomorphic embedding nor a bi-Lipschitz embedding into a separable dual Banach space. In 2007, Kalton proved that it does not coarsely embed into any reflexive Banach space. Whether $c_0$ can coarsely embed into a separable dual space has been a central question in coarse geometry of Banach spaces.
We first present some obstructions to coarse universality in separable dual spaces from our joint work with Cory Krause and Steve Jackson. In particular, we discuss a class of spaces that admit equi-coarse embeddings of Kalton’s interlacing graphs but are not coarsely universal. The proof uses Ramsey ultrafilters together with an absoluteness argument from set theory.
In the second part, we present a construction, discovered by AI, of a coarsely universal separable dual Banach space. This answers the coarse embedding question for $c_0$ affirmatively.
Abstract: Modern AI systems are increasingly built by placing trained models inside larger computational loops. Inference-time algorithms are a basic instance of this idea: they use one or more trained models at test time to incorporate new information, exploit pretrained models as priors, and trade computational effort for accuracy, sample quality, or control. Examples include generator-verifier search for reasoning, diffusion models for solving inverse problems, and reward-guided generation. Theoretically, this revisits a classical question from optimization and theoretical computer science: what can be done with access to an oracle? Here, however, the oracles are new and non-standard: they model the capabilities of large pretrained models, making them powerful, but also imperfect because they are learned. This combination leads to new questions about algorithm design and error propagation.
This talk studies two central aspects of this paradigm: computational efficiency and error propagation. The first vignette considers generator-verifier systems, and shows how stochastic backtracking can trade additional computation for accuracy, giving a principled version of test-time scaling even with imperfect learned oracles. The second vignette studies diffusion steering: when can we efficiently bias a pretrained diffusion model toward higher-reward samples while staying close to the original model? We show that tractability depends strongly on both the reward structure and the alignment objective, and that simple primitives—such as sampling from linear tilts—can be surprisingly useful for handling richer reward classes.
Speaker: Dmitry Jakobson (McGill University) - https://www.math.mcgill.ca/jakobson/ When: Tue, November 17, 2026 - 12:30pm Where: Kirwan Hall 1310
View Abstract
Abstract: We shall discuss conformal invariants that arise from nodal sets of certain eigenfunctions of conformally covariant operators, as well as applications to scalar curvature prescription problems. If time permits, we shall discuss related results in graph theory.
Speaker: Henri Berestycki (University of Maryland and EHESS Paris) - When: Tue, November 17, 2026 - 12:30pm Where: Kirwan Hall 3206
View Abstract
Abstract: This talk is about a predator-prey system where the predators are excluded from certain areas. Examples are Marine Protected Areas, or recently reported observations about how canids or felids occupy their territory. We investigate the effect of such a restriction of the predation area on both predators and prey. We especially analyze the influence of the extent and shape of the exclusion zone on the persistence and total size of the predator population. I report here on joint work with Bill FAGAN and Alex SAFSTEN.
Speaker: Dan Velleman (Amherst College) - When: Tue, November 17, 2026 - 3:30pm Where: Kirwan Hall 1311
View Abstract
Abstract: Lean is software that helps mathematicians write proofs. One of the most important ways in which it does this is that it checks the correctness of proofs. The underlying theory on which Lean is based is a version of dependent type theory called the calculus of inductive constructions. After a brief demonstration of Lean, I will introduce the calculus of inductive constructions and explain how Lean uses it to check proofs.
Speaker: Dongryul Kim (IAS) - https://dongryulkim.github.io/ When: Thu, November 19, 2026 - 2:00pm Where: Kirwan Hall 3206
View Abstract
Abstract: In this talk, we discuss recent developments of geometric methods to classify (infinite) invariant measures in various settings, homogeneous and non-homogeneous. We then discuss new counting results for objects on hyperbolic surfaces. This is joint work with Inhyeok Choi.
Speaker: Pedro Silva (Pedro Luis do Nascimento Silva (SCIENCE and National School of Statistical Sciences, Brazil)) - https://ence.academia.edu/PedroLuisdoNascimentoSilva?utm_source=chatgpt.com When: Thu, November 19, 2026 - 3:30pm Where: Kirwan Hall 1311
View Abstract
Abstract: Conducted annually since 2005, the NIC.br ICT Households Survey (ICTHS)
aims to map ICT access and usage in urban and rural households in Brazil,
covering individuals 10 years old or older. Redesign of the sample for this
survey was completed for the 2025 round and beyond. The redesign was
motivated by the need to incorporate the updated sampling frame provided by
the Brazilian Population Census 2022, as well as to enable improved precision
for state-level estimates. Optimization techniques were used to form and later to
stratify Primary Sampling Units (PSUs). Census Enumeration Areas (CEAs)
were combined to form the PSUs. The design is a stratified three stage cluster
sample, with PSUs stratified by geography and average income of heads of
household in the PSU. Sampling of PSUs and of CEAs within PSUs that have
more than two CEAs was carried out using Pareto PPS Sampling. Sampling of
households within each PSU or CEA is by simple random sampling without
replacement. I will highlight some challenges faced during the design and how
they were overcome with some of the design solutions.