Felix Otto (Max-Planck-Institut für Mathematik, Leipzig)
Stochastic homogenization abstract
Abstract:
In engineering applications, heterogeneous media are often described in statistical terms. This partial knowledge is sufficient to determine the effective, i.e. large-scale behavior. This effective behavior may be inferred from the Representative Volume Element (RVE) method. I report on last decades progress on the quantitative understanding of what is called stochastic homogenization of linear elliptic partial differential equations with random coefficient fields: optimal error estimates of the RVE method and the homogenization error, and the leading-order characterization of fluctuations. Methods connect to elliptic regularity theory, and in fact lead to a fresh look upon this classical area, and to concentration of measure arguments.In this course, I try to be both self-contained and efficient, by focussing on simple situations. Some prior knowledge on linear elliptic equations and functional analysis will be more helpful than advanced knowledge on probability theory, beyond elementary concepts like expectation or variance.
10:00 • online
Pedro Tamaroff (Trinity College Dublin)
Abstract:
The Hochschild–Kostant–Rosenberg theorem is a classical algebro-geometric result: Hochschild homology and cohomology of a smooth commutative algebra can be computed as the space of differential forms and of poly-vector fields on it, respectively. It says, in other words, how to compute the cohomology of a commutative algebra if we pull it back through the canonical projection of operads p : Ass –> Com and consider it merely to be associative. This result was then exploited to study the deformation theory of smooth commutative algebras, and obtain the celebrated Kontsevich formality theorem (cf. D. Tamarkin’s proof). In this talk, I will explain how to answer the following natural generalization of this question: given a map of operads f : P –> Q and a smooth Q-algebra A, how can one construct a space Ω*(A) of ‘differential forms for A’ and when can one produce an isomorphism from the (Hochschild) homology of the pullback P-algebra A to Ω*(A)? In particular, I’ll explain how to recover the usual Hochschild–Kostant–Rosenberg theorem using operadic homological algebra. This is joint work with Ricardo Campos (IMAG, Universite Montpellier, CNRS), arXiv:2010.08815.
10:15 • EPF Lausanne, Zoom Seminar
Jasmin Matz (University of Copenhagen)
Abstract:
Suppose M is a closed Riemannian manifold with an orthonormal basis B of L^2(M) consisting of Laplace eigenfunctions. A classical result of Shnirelman and others proves that if the geodesic flow on the cotangent bundle of M is ergodic, then M is quantum ergodic, in particular, on average, the probability measures defined by the functions f in B on M tends on average towards the Riemannian measure on M in the high energy limit (i.e, as the Laplace eigenvalues of f -> infinity).
We now want to look at a level aspect of this property, namely, instead of taking a fixed manifold and high energy eigenfunctions, we take a sequence of Benjamini-Schramm convergent compact Riemannian manifolds M_j together with Laplace eigenfunctions f whose eigenvalue varies in short intervals. This perspective has been recently studied in the context of graphs by Anantharaman and Le Masson, and for hyperbolic surfaces and manifolds by Abert, Bergeron, Le Masson, and Sahlsten. In my talk I want to discuss joint work with F. Brumley in which we study this question in higher rank, namely sequences of compact quotients of SL(n,R)/SO(n).
11:00 • Universität Basel, Zoom Seminar
Zhuo Jin (University of Melbourne, Australia)
Abstract:
This work studies a deep learning approach to find optimal insurance strategies for insurance companies. Due to the randomness of the financial ruin time to terminate the control processes, a Markov chain approximation-based iterative deep learning algorithm is developed to study this type of infinite-horizon optimal control problems. The optimal controls are approximated as deep neural networks. The framework of Markov chain approximation plays a key role in building the iterative equations and initialization of the algorithm. Optimal parameters of neural networks are then obtained iteratively. Convergence of the algorithm is studied. Satisfactory computation efficiency and accuracy are achieved as presented in numerical examples.
zoom link available upon request
11:00 • Université de Lausanne, Zoom Seminar
Marco Sutti (Université de Genève)
14:15 • Université de Genève, Section Mathématiques, Centre Acacias II, Salle 17
Dr. Alessandro Audrito (ETH Zürich, Switzerland)
The Neumann problem for some nonlocal elliptic operators: Regularity up to the boundary abstract
Abstract:
In this seminar, we will review some known results concerningthe regularity of solutions to some local and nonlocal ellipticequations, and then we will focus on recent developments about boundaryregularity of solutions to a class of "Neumann" nonlocal ellipticequations. The new results have been obtained in a joint work with J.-C. Felipe-Navarro (UPC), X. Ros-Oton (UB).
15:15 • Online via Zoom
Prof. Dr. Anna Wienhard (Heidelberg)
Growth of groups, entropy and Hausdorff dimension
16:30 • Zoom
Corentin Lena (Université de Neuchâtel)
Abstract:
Les modes de vibration stationnaires de systèmes physiques, tels des cordes, membranes ou cavités, sont décrits mathématiquement par les solutions de problèmes aux valeurs propres. Celles-ci donnent alors la fréquence et les fonctions propres l\'amplitude des vibrations. Dans un système de dimension 1, le nombre de zéros d\'une fonction propre (points immobiles) est donné par l\'indice de la valeur propre. Nous verrons dans quelle mesure ceci peut (ou pas) être étendu à des dimensions supérieures. Plus précisément, nous chercherons les fonctions propres, dites Courant-strictes, qui vérifient une propriété analogue. Je présenterai certains développements récents de ce problème très classique.
17:00 • Université de Neuchâtel, Zoom Seminar