Nguyen-Bac Dang (Université Orsay)
Abstract:
In this talk, based on a joint work with Charles Favre, I will explain one problem that arise in algebraic dynamics, concerning the growth under iteration of the algebraic degree of a given rational map.
We will see how this question can be related to the study of the spectral properties of certain pullback operators acting on a suitable Banach space of algebraic cycles.
14:15 • EPF Lausanne, room CM 0 9 and zoom
Marco De Renzi (Universität Zürich)
Représentations quantiques et homologiques des groupes de modulaires des surfaces abstract
Abstract:
La topologie quantique fournit des familles d\'invariants très organisées, dont la définition est flexible et générale. Pour certains d\'entre eux, une reformulation classique (plus précisément homologique) est connue. Ceci permet de garder le contrôle sur le contenu topologique des constructions résultantes, comme en témoigne la preuve spectaculaire de Bigelow de la linéarité des groupes de tresses. Pour le groupe modulaire Mod(Σ) d\'une surface Σ, nous expliquerons comment retrouver la famille de représentations quantiques associées au petit groupe quantique de sl(2) par une construction classique qui fait agir Mod(Σ) sur l\'homologie à coefficients tordus des espaces de configurations de Σ. Il s’agit d’un travail en cours, en collaboration avec Jules Martel.
15:00 • Université de Genève, Conseil Général 7-9, Room 1-15
Eugene Rabinovich (Notre Dame University)
Classical Bulk-Boundary Correspondences via Factorization Algebras abstract
Abstract:
A factorization algebra is a cosheaf-like local-to-global object which is meant to model the structure present in the observables of classical and quantum field theories.In the Batalin-Vilkovisky (BV) formalism, one finds that a factorization algebra of classical observables possesses, in addition to its factorization-algebraic structure, a compatible Poisson bracket of cohomological degree +1.Given a ``sufficiently nice\'\' such factorization algebra on a manifold $N$, one may associate to it a factorization algebra on $N\\times \\mathbb{R}_{\\geq 0}$.The aim of the talk is to explain the sense in which the latter factorization algebra ``knows all the classical data\'\' of the former.This is the bulk-boundary correspondence of the title.Time permitting, we will describe how such a correspondence appears in the deformation quantization of Poisson manifolds.
15:15 • ETH Zentrum, Rämistrasse 101, Zürich, Building HG, Room G 43
Prof. Dr. Evelyne Miot (University of Grenoble)
Dynamics of point vortices for the lake equations abstract
Abstract:
We study the asymptotic dynamics of point vortices for the lake equations with positive depth, when the vorticity is initially sharply concentrated around N points. More precisely, we show that the vorticity remains concentrated in some sense around N points for all times, and that the trajectories follow the level lines of the depth function. This is joint work with Lars Eric Hientzsch (Universität Bielefeld) and Christophe Lacave (Université Grenoble Alpes).
16:00 • ETH Zentrum, Rämistrasse 101, Zürich, Building HG, Room G 19.1
Elon Lindenstrauss (Hebrew University of Jerusalem)
Effective equidistribution of some unipotent flows with polynomial rates abstract
Abstract:
Joint work with Amir Mohammadi and Zhiren Wang.
A landmark result of Ratner gives that if $G$ is a real linear algebraic group, $\\Gamma$ a lattice in $G$ and if $u_t$ is a one-parameter unipotent subgroup of $G$, then for any $x \\in G/\\Gamma$ the orbit $u_t.x$ is equidistributed in a periodic orbit of some subgroup $L <G$, and moreover that the orbit of $x$ under $u_t$ is contained in this periodic $L$ orbit.
A key motivation behind Ratner\'s equidistribution theorem for one-parameter unipotent flows has been to establish Raghunathan\'s conjecture regarding the possible orbit closures of groups generated by one-parameter unipotent groups; using the equidistribution theorem Ratner proved that if $G$ and $\\Gamma$ are as above, and if $H<G$ is generated by one parameter unipotent groups then for any $x \\in G/\\Gamma$ one has that $\\overline{H.x}=L.x$ where $H < L < G$ and $L.x$ is periodic. Important special cases of Raghunathan\'s conjecture were proven earlier by Margulis and by Dani and Margulis by a different, more direct, approach.
These results have had many beautiful and unexpected applications in number theory, geometry and other areas. A key challenge has been to quantify and effectify these results. Beyond the case of actions of horospheric groups where there are several fully quantitative and effective results available, results in this direction have been few and far between. In particular, if $G$ is semisimple and $U$ is not horospheric no quantitative form of Ratner\'s equidistribution was known with any error rate, though there has been some progress on understanding quantitatively density properties of such flows with iterative logarithm error rates.
17:00 • Universität Basel, Zoom Seminar
Dr. Cristopher Salvi (Imperial College London)
From Neural SDEs to Neural SPDEs abstract
Abstract:
Many standard deep learning architectures may be interpreted as approximations to differential equations. The resulting continuous models are dubbed neural differential equations. ResNets is the approximation of ODEs while RNNs and variants (LSTM-GRU) are approximations of controlled differential equations (CDEs). In the first half of the talk, I will show how tools from rough path theory allows to understand properties of Neural CDEs to process time-varying data. Despite being an elegant solution to model temporal dynamics, Neural CDEs are not designed to process streams varying both in space and in time, such as videos. Stochastic PDEs (SPDEs) are the mathematical tool of choice to model many physical, biological, and economic systems subject to the influence of randomness. In the second half of the talk, I will present the Neural SPDE model capable of learning solution operators of PDEs with (possibly stochastic) forcing from partially observed data. Experiments on various semilinear SPDEs, including the stochastic Navier-Stokes equations, demonstrate how the Neural SPDE model is capable of learning complex spatiotemporal dynamics in a resolution-invariant way, with better accuracy and lighter training data requirements compared to alternative models, and up to 3 orders of magnitude faster than traditional solvers.
17:15 • ETH Zentrum, Rämistrasse 101, Zürich, Building HG, Room G 43
Prof. Dr. Jean-Michel Coron (Université Pierre et Marie Curie)
Stabilization of control systems: from the clepsydra to river regulation abstract
Abstract:
A control system is a dynamic system that can be acted upon using controls. For these systems, a fundamental problem is the question of stabilization: is it possible to stabilize a given unstable equilibrium using appropriate feedback laws? (Think of the classical experiment of a broomstick on the fingertip.) On this problem, we present some old devices and pioneering works (Ctesibius, Watt, Maxwell, Lyapunov...), more recent results, and an application to the regulation of the rivers La Sambre and La Meuse. The focus is on the positive or negative effects of non-linearities.
17:15 • EPF Lausanne, Salle MA B1 11
Prof. Dr. Agostino Capponi (Columbia University)
Goal Based Investment Management abstract
Abstract:
We develop a continuous time framework for goals-based investing, where the objective is to maximize the priority-weighted fundedness of the clients\' goals. We show that the value function of the control problem can be recovered as the unique viscosity solution to a nonlinear Hamilton-Jacobi-Bellman equation. Our analysis highlights the fundamental tradeoff between immediate goal consumption versus savings toward future goal liabilities. We show that it is optimal to fund an expiring goal up to the level where the marginal benefit of consuming wealth is exceeded by the marginal cost of reducing the fundedness of future goals. This tradeoff depends crucially on the interplay between goal amounts, goal deadlines, client\'s initial wealth and guaranteed income stream. Our comparative statics analysis reveals that the risk aversion towards a goal increases if the goal\'s deadline is approaching or if the priority of a goal relative to future goals increases. (joint work with Yuchong Zhang).
18:15 • ETH Zentrum, Rämistrasse 101, Zürich, Building HG, Room G 43