Michael Ehrig (Beijing Institute of Technology)
Lie Superalgebras via Schur-Weyl duality and Categorification abstract
Abstract:
In this talk, I will outline an approach to understand and describe the category of finite dimensional representations of a classical Lie superalgebra. Due to the non semi simplicity, methods different from the ones for semi-simple Lie algebras need to be applied to describe this category. Using variations of Schur-Weyl duality, respectively the fundamental theorems of invariant theory, we formulate the problem of understanding it in terms of centralizer algebras. These centralizer algebras are then described via methods from categorification of quantum groups and link invariants, yielding the description of the category of finite dimensional representations for some of the classical Lie superalgebras. This is joint work with Catharina Stroppel.
10:15 • UZH Irchel, Winterthurerstrasse 190, Zürich, Building Y27, Room H 28
Daniele Tantari (Scuola Normale Superiore)
Direct/Inverse Hopfield model and Restricted Boltzmann Machines abstract
Abstract:
Mean-field methods fail to reconstruct the parameters of the model when the dataset is clusterized. This situation is found at low temperatures because of the emergence of multiple thermodynamic states. The paradigmatic Hopfield model is considered in a teacher-student scenario as a problem of unsupervised learning with Restricted Boltzmann Machines (RBM). For different choices of the priors on units and weights, the replica symmetric phase diagram of random RBM’s is analyzed and in particular the paramagnetic phase boundary is presented as directly related to the optimal size of the training set necessary for a good generalization. The connection between the direct and inverse problem is pointed out by showing that inference can be efficiently performed by suitably adapting both standard learning techniques and standard approaches to the direct problem.
11:00 • Universität Basel, Fachbereich Mathematik, Spiegelgasse 1, Seminar Room 05.003
Daniele Tantari (Scuola Normale Superiore)
Abstract:Mean-field methods fail to reconstruct the parameters of the model when the dataset is clusterized. This situation is found at low temperatures because of the emergence of multiple thermodynamic states. The paradigmatic Hopfield model is considered in a teacher-student scenario as a problem of unsupervised learning with Restricted Boltzmann Machines (RBM). For different choices of the priors on units and weights, the replica symmetric phase diagram of random RBM\'s is analyzed and in particular the paramagnetic phase boundary is presented as directly related to the optimal size of the training set necessary for a good generalization. The connection between the direct and inverse problem is pointed out by showing that inference can be efficiently performed by suitably adapting both standard learning techniques and standard approaches to the direct problem.
https://probability.dmi.unibas.ch/seminar.html 11:00 • Universität Basel, Spiegelgasse 5, Room 00.003
Andrew Mathas (University of Sydney)
An overview of the representation theory of the symmetric groups abstract
Abstract:
A representation of the symmetric group is a vector space together with an action of the group. Studying these representations is a rich beautiful subject, with its roots in combinatorial and with connections to areas ranging from general linear groups, to Kac-Moody algebras, 2-category theory, knot theory, statistical mechanics and Markov traces. This talk will start with results from the classical representation theory and a summary of the main open problems in the field. We then slowly build towards some recent exciting developments with the aim of describing the Ariki-Brundan-Kleshchev categorification theorem. Roughly speaking, this result shows that the representation theory of the symmetric groups is intimately connected with the highest weight representation theory of the affine special linear groups.
13:00 • ETH Zentrum, Building Y23, Room G 04
Bill Duke (University of California, Los Angeles)
Topics in modern analytic number theory abstract
Abstract:
This series of lectures will treat a number of special topics all of which involve some aspect of the modular group SL(2,Z).
Many have as their natural starting point the class number formulas of Dirichlet for binary quadratic form or, equivalently, ideal classes of quadratic number fields. Most involve L-functions of various kinds and modular forms and their generalizations. These topics include:
New class number formulas, both for quadratic forms and higher degree forms, for example binary cubic form Properties of geometric objects associated to the modular group, including cm points, closed geodesics and infinite volume surfaces having closed geodesics as boundaries. The interaction between flat metrics and the hyperbolic metric, for instance the curvature of closed geodesics with respect to a flat metric. Quasimorphisms of the modular group and modular cocycles. Linking numbers of modular knots. Markov forms and modular billiards Connections with Diophantine approximation and the geometry of numbers.
I will attempt to provide enough background and historical context for each topic to make them accessible to non-experts, but hope to give enough "new" material to stimulate the interest of the experts as well. I will also give suggestions about areas for further research.
13:15 • ETH Zentrum, Rämistrasse 101, Zürich, Building HG, Room G 43
Prof. Dr. David Holmes (University of Leiden)
Tropical divisors and the double ramification cycle abstract
Abstract:
fter recalling some basics on logarithmic curves, we will explain in some detail a construction of Marcus and Wise of the stack of tropical divisors on a family of log curves. We will then explain how to see this as an incarnation of the double ramification cycle on the Picard stack over the stack of all prestable curves.
13:30 • ETH Zentrum, Rämistrasse 101, Zürich, Building HG, Room F 26.1
Dr. Michela Ceria (Università degli Studi di Milano)
Half error locator polynomials for efficient decoding of binary cyclic codes abstract
Abstract:
In 1990 Cooper suggested to use Groebner bases\' computations to decode cyclic codes and his idea gave rise to many research papers. In particular, as proved by Sala-Orsini, once defined the polynomial ring whose variables are the syndromes, the locations and the error values and considered the syndrome ideal, only one polynomial of a lexicographical Groebner basis for such ideal is necessary to decode (the general error locator polynomial - GELP). The decoding procedure only consists in evaluating this polynomial in the syndromes and computing its roots: the roots are indeed the error locations. A possible bottleneck in this procedure may be the evaluation part, since a priori the GELP may be dense. In this talk, focusing on binary cyclic codes with length n | 2^m - 1, correcting up to two errors, we give a Groebner-free, sparse analog of the GELP, the half error locator polynomial (HELP). In particular, we show that it is not necessary to compute the whole Groebner basis for getting such kind of locator polynomial and we construct the HELP, studying the quotient algebra of the polynomial ring modulo the syndrome ideal by a combinatorial point of view. The HELP turns out to have linear growth in the length n of the code: O((n+1)/2).
15:00 • Uni Neuchatel, B217
Ivan Cherednik (Lomonossow-Universität, Russland)
Riemann Hypothesis for Zeta-functions of Plane Curve Singularities abstract
Abstract:
Zeta-functions of plane curve singularities will be defined from scratch and discussed, including basic examples. The functional equation for them holds, however the Riemann Hypotheis fails for the corresponding L-functions in contrast to the smooth projective curves (Weil, Deligne). This can be fixed for sufficiently small q. Classically q is the cardinality of a finite field, but it becomes an arbitrary parameter in the new vintage of the theory, inspired by the connections with motivic and DAHA superpolynomials, and with Khovanov-Rozansky ones. The motivic superpolynomials will be fully defined; their relation to affine Springer fibers will be discussed at the end. Only very basic knowledge of rings/modules is assumed and no theory of curves and their singularities will be used.
15:15 • EPF Lausanne, Salle GC A3 30
Alexander Lytchak (Universität Köln)
Topological stability in upper curvature bounds abstract
Abstract:
In the talk I would like to discuss an analog of Pereleman\'s stability theorem in upper curvature bounds and related results about the topological regularity of spaces with upper curvature bounds and extendable geodesics. The talk is based on joint works with Koichi Nagano.
15:45 • ETH Zentrum, Rämistrasse 101, Zürich, Building HG, Room G 43
Nikhil Bansal (University of Eindhoven)
Online vector balancing abstract
Abstract:
Consider the setting where vectors in [-1,1]^n arrive online, and upon the arrival of vector v_t at time t, it must be immediately and irrevocably assigned a sign +1 or -1. The goal is to pick these signs to ensure that the norm of the signed sum of these vectors stays small. Such online discrepancy games were originally considered by Spencer in the 70\'s, but have received renewed attention due to applications in online algorithms.
I will give an overview of the results and techniques in the area and describe several new recent results that obtain close to optimal bounds in many settings. I will also discuss several basic questions that remain completely open.
Based on joint works with Joel Spencer, Haotian Jiang, Sahil Singla and Makrand Sinha.
16:15 • EPF Lausanne, Salle BC02
Gianira Nicoletta Alfarano (Universität Zürich)
New Algebraic Construction of MDP Convolutional Codes abstract
Abstract:
Convolutional codes were introduced in 1955 by Peter Elias, as an interesting natural generalization of classical block codes and they have been widely implemented in practice with applications in mobile communications, satellite communications, and data streaming. Unfortunately, there exist very few algebraic general constructions of convolutional codes and most of the existing ones that have good designed distance have been found by computer search. The algebraic construction of convolutional codes with a large distance that admits an efficient decoding algorithm by exploiting the algebraic structure of the code remains a big open problem in the area. In this talk, we introduce a new algebraic construction of a class of convolutional codes. Since these codes are built upon generalized Vandermonde matrices, they can be seen as an extension of Reed-Solomon (RS) block codes to the context of convolutional codes. For this reason, we will refer to them as Weighted RS convolutional codes. We will discuss the MDP property of these codes and we will give some asymptotic results on the field size that they require to be built. Finally, we will give a comparison with the existing constructions.
16:15 • Uni Neuchatel, B217
Prof. Dr. Steffen Börm (Christian-Albrechts-Universität zu Kiel)
Fast Boundary Element Methods for Electrostatic Field Computations abstract
Abstract:
We consider the computation of electrostatic potentials by the boundary element method. In order to obtain O(h²) convergence of discrete solutions, we have to employ piecewise linear basis functions and piecewise quadratic parametrizations of the surface.Constructing the data-sparse approximations of the integral operators required for high accuracies poses several challenges: methods like ACA or GCA require the computation of individual matrix entries, and since the supports of basis functions are spread across multiple triangles, this computation is far more computationally expensive than for simple discontinuous basis functions. Alternative techniques like HCA allow us to significantly reduce the computational work.Another challenge is the parametrization of the curved triangles: the Gramian and the normal vector are no longer constant on each triangle, but have to be computed in each quadrature point, and this increases the necessary work even further.Combining efficient quadrature techniques with HCA matrix compression, algebraic coarsening and recompression, and Krylov solvers allows us to handle surface meshes with up to 18 million triangles on relatively affordable servers while preserving the theoretically predicted convergence rate of the underlying discretization scheme.
16:15 • ETH Zentrum, Rämistrasse 101, Zürich, Building HG, Room E 1.2
Christophe Garban (Université de Lyon)
Inverted orbits of exclusion processes, diffuse-extensive- amenability and (non-?)amenability of the interval exchanges abstract
Abstract:
After a brief introduction on the notion of amenability for groups, I will focus on the group of interval exchanges (the IET group) which is believed to be amenable. One of the (many) equivalent criteria to showthat a group is amenable is Kesten\'s criterion on the return probabilities of random walks. In the case of G=IET, a recent work by Juschenko, Matte Bon, Monod and De La Salle provides a new criterion which is also of probabilistic nature. This new criterion involves the size of the inverted orbit of a certain random walk on the wobbling group W(Z^d) of permutations of Z^d. The aim of this talk will be to introduce natural models of random walks on permutations of Z^d for which this criterion can be analyzed. The talk will not require any prerequisites.
17:15 • ETH Zentrum, Rämistrasse 101, Zürich, Building HG, Room G 43