Immanuel van Santen (Universität Basel)
Embeddings of smooth affine varieties into linear algebraic groups abstract
Abstract:
This is joint work with Peter Feller. There are the following fundamental questions concerning embeddings from an object Z into another object X:
1. (Existence) Does there exists an embedding of Z into X?
2. (Uniqueness) Having two embeddings f, g of Z into X, does there exists an automorphism ψ of X such that g = ψ ◦ f?
In this talk, we will mainly focus on the first question in the category of affine varieties, where Z is smooth and X is alinear algebraic group. Amongst other things, we will discuss the following result.
Theorem. For every simple linear algebraic group G and every smooth affine variety Z with dim G > 2 dim Z + 1, there exists an embedding of Z into G.
The proof is based upon parametric transversality results for flexible affine varieties due to Kaliman. We will also discuss the following result.
Proposition. For every non-finite algebraic group G and every d ≥ dim G / 2, there exists an irreducible smooth affine variety of dimension d that does not admit an embedding into G.
This proposition implies the optimality of the above existence theorem up to a possible improvement of the dimension bound by one. It’s proof is an adaptation of a Chow-group-based argument due to Bloch, Murthy, and Szpiro for the affine space.
10:45 • EPF Lausanne, Salle MA B1 11
Romain Tessera (Université Pierre et Marie Curie, Paris Diderot)
Which groups coarsely embed into a hyperbolic space? - Séminaire Mini course on Geometric Group Theory abstract
Abstract:
An amenable group admits a coarse embedding into a hyperbolic space if and only if it is virtually nilpotent.This result is proved by a combination of techniques involving analysis (Poincaré profiles) and ergodic theory (measured subgroups).
13:15 • Université de Genève, Conseil Général 7-9, Room 1-05
Arthur Forey (ETH Zürich)
Equidistribution of exponential sums on abelian varieties abstract
Abstract:
Many exponential sums over finite fields, such as Gauss or Salié-Kloosterman sums, appear as the Fourier-Melin transform of the trace function of an l-adic sheaf on a commutative algebraic group. We are interested in the equidistribution of such sums as the character varies. Generalizing work by Deligne and Katz in the cases of additive and multiplicative groups, a Tannakian formalism always controls the equidistribution. In this talk, I will illustrate this result in some concrete cases where the group is an abelian variety. This is a collaboration with Javier Fresán and Emmanuel Kowalski.
13:15 • EPF Lausanne, Salle tba
Jean-Marc Deshouillers (Université de Bordeaux)
Abstract:
https://drive.google.com/file/d/138X5V4kOI0rBf1kHiP3iszyjpePh2VKk/view
17:00 • Universität Basel, Zoom Seminar
Christer Oscar Kiselman
Discrete phenomena, which are of growing interest in science, can be difficult but need to be understood abstract
Abstract:
Mathematical models using real or complex numbers have been most successful. Discrete objects, like mosaics and carpets, have been around during millennia. With the appearance of computers and digital cameras, they are now of growing importance and are subject to intensive research. I shall compare Euclidean geometry with the much younger digital geometry, and try to exhibit their similarities and differences.
17:00 • Université de Fribourg, room Physiology 1.100
Dr. Anastasis Kratsios (University of Basel)
Universal Approximation under Constraints is Possible with Transformers abstract
Abstract:
Many practical problems need the output of a machine learning model to satisfy a set of constraints, K. Nevertheless, there is no known guarantee that classical neural network architectures can exactly encode constraints while simultaneously achieving universality. We provide a quantitative constrained universal approximation theorem which guarantees that for any non-convex compact set K and any continuous function f:Rn→K, there is a probabilistic transformer F whose randomized outputs all lie in K and whose expected output uniformly approximates f. Our second main result is a "deep neural version" of Berge\'s Maximum Theorem (1963). The result guarantees that given an objective function L, a constraint set K, and a family of soft constraint sets, there is a probabilistic transformer F that approximately minimizes L and whose outputs belong to K; moreover, F approximately satisfies the soft constraints. Our results imply the first universal approximation theorem for classical transformers with exact convex constraint satisfaction. They also yield that a chart-free universal approximation theorem for Riemannian manifold-valued functions subject to suitable geodesically convex constraints.Joint work with Behnoosh Zamanlooy, Ivan Dokmanić and Tianlin Liu.
17:15 • ETH Zentrum, Rämistrasse 101, Zürich, Building HG, Room G 43