Upcoming Events

go to archive

Wednesday, 25.04.2018, 10:00 (WIAS-ESH)
Forschungsseminar Mathematische Statistik
N. Baldin, University of Cambridge, GB:
Optimal link prediction with matrix logistic regression
more ... Location
Weierstraß-Institut, Mohrenstr. 39, 10117 Berlin, Erdgeschoss, Erhard-Schmidt-Hörsaal

Abstract
In this talk, we will consider the problem of link prediction, based on partial observation of a large network, and on side information associated to its vertices. The generative model is formulated as a matrix logistic regression. The performance of the model is analysed in a high-dimensional regime under a structural assumption. The minimax rate for the Frobenius-norm risk is established and a combinatorial estimator based on the penalised maximum likelihood approach is shown to achieve it. Furthermore, it is shown that this rate cannot be attained by any (randomised) algorithm computable in polynomial time under a computational complexity assumption. (joint work with Q. Berthet)

Further Informations
Forschungsseminar “Mathematische Statistik”

Host
Humboldt-Universität zu Berlin
Universität Potsdam
WIAS Berlin
Thursday, 03.05.2018, 14:00 (WIAS-ESH)
Seminar Numerische Mathematik
Dr. H. Stephan, WIAS Berlin:
One million perrin pseudo primes including a few giants
more ... Location
Weierstraß-Institut, Mohrenstr. 39, 10117 Berlin, Erdgeschoss, Erhard-Schmidt-Hörsaal

Abstract
Pseudoprimes are integers that are no primes but behave like them in some sense. Suppose we have a theorem like the following: If n is a prime, then statement A(n) holds. In general, the opposite is not true: It may be that A(n) holds, but n is a composite number, a so-called pseudo prime with respect to statement A. Pseudoprimes are interesting if they are very rare, as for instance Perrin's pseudoprimes, the smallest of which is 271441. The talk introduces pseudoprimes which are based on recurrent sequences. In addition, some new numerical results on Perrin's pseudoprimes and a fast algorithm for their calculation are presented.

Host
WIAS Berlin
Thursday, 03.05.2018, 15:00 (WIAS-ESH)
Seminar Numerische Mathematik
L. Feierabend, Universität Duisburg-Essen:
Model development for flowing slurry electrodes in zinc-air batteries
more ... Location
Weierstraß-Institut, Mohrenstr. 39, 10117 Berlin, Erdgeschoss, Erhard-Schmidt-Hörsaal

Abstract
Slurry or suspension electrodes are gaining a renewed interest for large-scale energy storage technologies due to potentially higher energy densities compared to redox flow batteries [1]. A suspension electrode typically consists of a liquid electrolyte and a solid material, which is chemically active and electrically conducting. For the investigated zinc-air flow batteries employing a suspension electrode, microscopic zinc particles are suspended in an aqueous potassium hydroxide electrolyte [2, 3]. Sedimentation of the metallic zinc is minimized by adding a gelling agent to the electrolyte. The gelling agent and the high particle loading lead to a pseudo-plastic rheological behavior. When the flowable suspension electrode is pumped through the channels of the zinc-air fuel cell setup, the dynamic percolation network within the suspension dictates the active electrode surface area and the maximum discharge power density. Therefore, it is desirable to investigate the influence of the local flow conditions on the particle percolation and consequently the electrochemical performance of the battery cell with adequate simulation methods. In the developed three-dimensional numerical model for the flowing suspension electrode, the complex, non-Newtonian two-phase flow is approximated by a coupling of an Eulerian continuum description for the electrolyte and a discrete, Lagrangian particle description for the motion and interaction of the microscopic particles. The partial differential equations for momentum, species, charge and energy are discretized by the finite volume method and implemented in the OpenFOAM library [4]. The particle motion including multiple simultaneous particle contacts is described with the discrete element method using the LIGGGHTS library [5]. Coupling between the particle and fluid phase is realized with the CFD-DEM method using the CFDEM library [5], where an empirical description accounts for the momentum exchange between the non-Newtonian fluid and the densely distributed particle assemblies. Due to relatively high particle concentrations, the volume displacement by the particles is considered in the electrolyte continuum model. A half-cell model for the anode part of the zinc-air flow battery is implemented, which accounts for the flow characteristics via the described CFD-DEM coupling method. Simultaneously, the charge and species transport is considered according to the classic porous electrode theory by Newman [6] adapted with newer formulations for concentrated electrolytes [7]. In contrast to electrodes with a static porous matrix, the heterogeneous local porosities change temporally depending on the evolving particle distributions. Additionally, the active electrode surface area in each finite volume is dependent on the percolation network from the considered local point to the current collector surface. To estimate the basic parameters for the charge transfer and species transport, a flat-plate electrode with a flowing electrolyte without particles is investigated. In this rather simple reference case, discontinuities of the current density close to the electrode-electrolyte-interface could be observed, which should be inherently excluded due to the model formulations. These inconsistencies will be discussed in detail. Funding by the Federal Ministry for Economic Affairs and Energy (BMWi) of the project â?œZnPLUS -Wiederaufladbare Zink-Luft-Batterien zur Energiespeicherungâ?? and the project â?œZnMobil - Mechanisch und elektrisch wiederaufladbare Zink-Luft-Batterie für automobile Anwendungenâ?? is gratefully acknowledged.

Host
WIAS Berlin
Wednesday, 09.05.2018, 15:15 (WIAS-ESH)
Berliner Oberseminar „Nichtlineare partielle Differentialgleichungen” (Langenbach-Seminar)
Prof. A. Jüngel, Technische Universität Wien, Österreich:
Cross-diffusion systems with entropy structure
more ... Location
Weierstraß-Institut, Mohrenstr. 39, 10117 Berlin, Erdgeschoss, Erhard-Schmidt-Hörsaal

Abstract
here

Further Informations
Berliner Oberseminar “Nichtlineare Partielle Differentialgleichungen” (Langenbach Seminar)

Host
WIAS Berlin
Humboldt-Universität zu Berlin
Wednesday, 16.05.2018, 15:15 (WIAS-ESH)
Berliner Oberseminar „Nichtlineare partielle Differentialgleichungen” (Langenbach-Seminar)
Prof. A. Mielke, WIAS Berlin:
Finding limiting dissipative potentials via EDP convergence
more ... Location
Weierstraß-Institut, Mohrenstr. 39, 10117 Berlin, Erdgeschoss, Erhard-Schmidt-Hörsaal

Abstract
A gradient structure for an evolutionary system is a triple, a so-called gradient system, consisting of a state space, a possibly time-dependent energy functional, and a dissipation potential that induces the given evolutionary system. The gradient structure, which is not unique, contains additional thermodynamical information, e.g. on the fluctuations in an associated microscopic model. Considering a family of gradient systems depending on a small parameter, it is natural to ask for the limiting or effective gradient system if the parameter tends to 0. We propose some ideas towards the derivation of the effective gradient structure that are based on De Giorgi's Energy-Dissipation Principle (EDP). We discuss several versions of EDP convergence and show by examples that the theory is flexible enough to allow situations where starting from quadratic dissipation potentials we arrive at effective dissipation potentials that are no longer quadratic.

Further Informations
Berliner Oberseminar “Nichtlineare Partielle Differentialgleichungen” (Langenbach Seminar)

Host
WIAS Berlin
Humboldt-Universität zu Berlin
Thursday, 17.05.2018, 10:15 (WIAS-406)
Seminar Nichtlineare Optimierung und Inverse Probleme
C. Geiersbach, Universität Wien, Österreich:
A projected stochastic gradient algorithm for optimization with random elliptic PDE constraints
more ... Location
Weierstraß-Institut, Mohrenstr. 39, 10117 Berlin, 4. Etage, Weierstraß-Hörsaal (Raum: 406)

Host
WIAS Berlin
Wednesday, 23.05.2018, 15:15 (WIAS-ESH)
Berliner Oberseminar „Nichtlineare partielle Differentialgleichungen” (Langenbach-Seminar)
Prof. Dr. P. Colli, Università di Pavia, Italien:
A Cahn--Hilliard system with convection and dynamic boundary conditions: Well-posedness and optimal velocity control
more ... Location
Weierstraß-Institut, Mohrenstr. 39, 10117 Berlin, Erdgeschoss, Erhard-Schmidt-Hörsaal

Further Informations
Berliner Oberseminar “Nichtlineare Partielle Differentialgleichungen” (Langenbach Seminar)

Host
WIAS Berlin
Humboldt-Universität zu Berlin
Wednesday, 30.05.2018, 10:00 (WIAS-ESH)
Forschungsseminar Mathematische Statistik
F. Schäfer, California Institute of Technology, USA:
Compression, inversion, and approximate PCA of dense kernel matrices at near-linear computational complexity
more ... Location
Weierstraß-Institut, Mohrenstr. 39, 10117 Berlin, Erdgeschoss, Erhard-Schmidt-Hörsaal

Abstract
Many popular methods in machine learning, statistics, and uncertainty quantification rely on priors given by smooth Gaussian processes, like those obtained from the Mat Ì?ern covariance functions. Furthermore, many physical systems are described in terms of elliptic partial differential equa- tions. Therefore, implicitely or explicitely, numerical simulation of these systems requires an efficient numerical representation of the correspond- ing Greenâ??s operator. The resulting kernel matrices are typically dense, leading to (often prohibitive) O N2 or O N3 computational complexity. In this work, we prove rigorously that the dense N Ã? N kernel matri- ces obtained from elliptic boundary value problems and measurement points distributed approximately uniformly in a d-dimensional domain can be Cholesky factorised to accuracy ε in computational complexity O N log2(N)log2d(N/ε) in time and O N log(N)logd(N/ε) in space. For the closely related Mat Ì?ern covariances we observe very good results in practise, even for parameters corresponding to non-integer order equa- tions. As a byproduct, we obtain a sparse PCA with near-optimal low- rank approximation property and a fast solver for elliptic PDE. We emphasise that our algorithm requires no analytic expression for the covariance function. Our work is inspired by the probabilistic interpretation of the Cholesky factorisation, the screening effect in spatial statistics, and recent results in numerical homogenisation.

Further Informations
Forschungsseminar “Mathematische Statistik”

Host
Humboldt-Universität zu Berlin
Universität Potsdam
WIAS Berlin