Upcoming Events

Many events are currently organized online. Information on how to access these events can be found by clicking “more” below the respective entry.


go to archive

Tuesday, 07.02.2023, 15:00 (WIAS-405-406)
Seminar Modern Methods in Applied Stochastics and Nonparametric Statistics
PhD Alain Rossier, University of Oxford, GB:
Asymptotic analysis of deep residual networks (hybrid talk)
more ... Location
Weierstraß-Institut, Mohrenstr. 39, 10117 Berlin, 4. Etage, Raum: 405/406

Abstract
Residual networks (ResNets) have displayed impressive results in pattern recognition and, recently, have garnered considerable theoretical interest due to a perceived link with neural ordinary differential equations (neural ODEs). This link relies on the convergence of network weights to a smooth function as the number of layers increases. We investigate the properties of weights trained by stochastic gradient descent and their scaling with network depth through detailed numerical experiments. We observe the existence of scaling regimes markedly different from those assumed in neural ODE literature. Depending on certain features of the network architecture, such as the smoothness of the activation function, one may obtain an alternative ODE limit, a stochastic differential equation (SDE) or neither of these. Furthermore, we are able to formally prove the linear convergence of gradient descent to a global optimum for the training of deep residual networks with constant layer width and smooth activation function. We further prove that if the trained weights, as a function of the layer index, admit a scaling limit as the depth increases, then the limit has finite 2-variation.

Further Informations
Dieser Vortrag findet bei Zoom statt: https://zoom.us/j/492088715

Host
WIAS Berlin
Wednesday, 08.02.2023, 10:00 (WIAS-ESH)
Forschungsseminar Mathematische Statistik
Prof. Dr. Vanessa Didelez, Universität Bremen:
Causal reasoning and causal discovery with applications in epidemiology
more ... Location
Weierstraß-Institut, Mohrenstr. 39, 10117 Berlin, Erdgeschoss, Erhard-Schmidt-Hörsaal

Abstract
Many data analyses ultimately aim at answering causal research questions: We may want to assess and quantify the potential effects of certain decisions, interventions or policies, e.g. will a sugar tax or more playgrounds reduce childhood obesity? Will participation in a special training programme for the unemployed increase the chances of finding employment? Is a national mammography screening programme actually helpful in preventing deaths from breast cancer? Such questions are about causal relations and go beyond mere prediction; indeed, methods that are optimised for prediction will often give biased results for causal targets. Especially when we use non-experimental, i.e. observational data to try and answer questions about causal relations, tailored methods relying on specific assumptions are called for. The talk will review the main concepts, fundamental assumptions and basic principles for causal learning and focus on methods of causal discovery (aka structure learning). The latter have their roots in probabilistic approaches to artificial intelligence (AI) and, together with broader methods of causal inference in general, have recently seen a great revival in AI. This increased activity might be due to the realization "that many hard open problems of machine learning and AI are intrinsically related to causality" (Schölkopf, 2019). However, applications in epidemiology still pose a number of practical challenges; these include, for instance, handling incomplete, mixed, heterogenous and temporal data. I will illustrate some of the methods, their issues and proposed solutions with the analysis of a children's cohort data.

Further Informations
Der Vortrag findet bei Zoom statt: https://zoom.us/j/159082384

Host
Humboldt-Universität zu Berlin
Universität Potsdam
WIAS Berlin
Wednesday, 08.02.2023, 11:30 (WIAS-405-406)
Seminar Interacting Random Systems
Dave Jacobi, TU Berlin:
Super-Brownian motion with dormancy
more ... Location
Weierstraß-Institut, Mohrenstr. 39, 10117 Berlin, 4. Etage, Raum: 405/406

Abstract
The majority of species exhibit a behaviour called Dormancy, in which the individuals switch into a state of low metabolic activity, that protects them from harsh environmental conditions and in this way increases their chance of survival. Therefore models from mathematical population biology have to incorporate this phenomenon. We will extend the classical Super-Brownian motion, which is a measure-valued branching Markov process, to model Dormancy and retrieve a process that we will call on/off Super-Brownian motion. This process has many interesting properties that are often closely related to classical Super-Brownian motion, but at the same time exhibit new and different behaviour. We will go on a round trip of these properties. Also, if time permits, we will look at the close relation between the total mass process of superprocesses and an excursion process, which can help us understand the genealogy of these measure-valued processes.

Further Informations
Seminar Interacting Random Systems (Hybrid Event)

Host
WIAS Berlin
Wednesday, 08.02.2023, 15:15 (WIAS-ESH)
Berliner Oberseminar „Nichtlineare partielle Differentialgleichungen” (Langenbach-Seminar)
Jun. Prof. Dr. Patrick Tolksdorf, Johannes Gutenberg-Universität Mainz:
Lp-extrapolation of the generalized Stokes operator (hybrid talk)
more ... Location
Weierstraß-Institut, Mohrenstr. 39, 10117 Berlin, Erdgeschoss, Erhard-Schmidt-Hörsaal

Abstract
Please see here.

Further Informations
Hybridveranstaltung - Teilnahme vor Ort bitte bei Dr. A. Glitzky (annegret.glitzky@wias-berlin.de) anmelden.
Hybrid event - please give Dr. A. Glitzky (annegret.glitzky@wias-berlin.de) notice of your on-site participation.

Host
Humboldt-Universität zu Berlin
WIAS Berlin