Events

EU Regional School - Fonnesbeck Seminar

Location: AICES Seminar Room 115, 1st floor, Schinkelstr. 2, 52062 Aachen

Prof. Christopher Fonnesbeck, Ph.D. - Probabilistic Programming with Python

Department of Biostatistics, Vanderbilt University Medical Center, USA

Abstract

This intermediate-level course will provide students with hands-on experience applying practical Bayesian statistical modeling methods on real data. PyMC3 is a high-level Python library for building statistical models using probabilistic programming, and fitting them using modern Bayesian computational methods. I will provide an introduction to Bayesian inference and prediction, followed by a tutorial on probabilistic programming with PyMC3, including the use of Markov chain Monte Carlo (MCMC) and Variational Inference (VI), using real-world datasets. The last part of the course will focus on modeling strategies and how to avoid various pitfalls when applying Bayesian statistics to your own work. The course will assume familiarity with Python and with basic statistics (e.g. probability), but does not require previous experience with Bayesian methods or probabilistic programming.

I³MS - Saxena Seminar

Location: AICES Seminar Room 115, 1st floor, Schinkelstr. 2, 52062 Aachen

Prof. Dr. Anupam Saxena - Contact in Topology Optimization: Challenges and Solution Strategies

Department of Mechanical Engineering, Indian Institute of Technology, Kanpur

Abstract

Topology Optimization of large deformation planar continua will be
formulated. Two parameterization schemes, namely, frame- and continuum-
based, will be exemplified using small deformation theory for familiarity
with existing, gradient- and function- based, topology optimization
schemes. Within continuum parameterization, rectangular and hexagonal
tessellations will be discussed. The first set of challenges pertaining to
connectivity singularities, namely, checkerboards and point flexures will
be highlighted, and remedies will be mentioned. Constituents of planar
continua undergoing large deformation have a tendency to come in contact.
This aspect can either be avoided, or used to advantage
to design continua that can deliver the desired, intricate deformation
characteristics. Many challenges, when adapting topology optimization to
incorporate contact interactions, will be highlighted and solution
strategies will be discussed. These challenges pertain to mesh handling,
generation/evaluation of binary continua, non-convergence in finite
element analysis, generation of rigid contact surfaces in vicinity of
largely deforming constituents, and other factors. It will be argued that
zero-order searches, albeit computationally costly, seem more viable in
addressing these problems. The talk will culminate with examples on large
deformation monolithic gripper-manipulator, and path generating continua
that can exhibit one or more desired kinks in the path by effectively
employing contact interactions. Some open problems will also be
highlighted.

EU Regional School - Püschel Seminar

Location: AICES Seminar Room 115, 1st floor, Schinkelstr. 2, 52062 Aachen

Prof. Dr. Markus Püschel - Optimal Performance Numerical Code: Challenges and Solutions

Department of Computer Science, ETH Zürich, Switzerland

Abstract

TBA

I³MS - Voigt Seminar

Location: AICES Seminar Room 115, 1st floor, Schinkelstr. 2, 52062 Aachen

Prof. Dr. Axel Voigt - Thin Films of Liquid Crystals: Modeling and Numerics

Institute of Scientific Computing, Technical University Dresden

Abstract

We consider a thin film limit of a Frank-Oseen and a Landau-de Gennes model. In the limiting process we observe a continuous transition where the normal and tangential parts of the director field and the Q-tensor decouple and various intrinsic and extrinsic contributions emerge. Main properties of the thin film model, like uniaxiality and parameter phase space, are preserved in the limiting process. For the derived surface Frank-Oseen and surface Landau-de Gennes model, we consider an $L^2$-gradient flow. The resulting vector- and tensor-valued surface partial differential equations are numerically solved to demonstrate realizations of the tight coupling of elastic and bulk free energy with geometric properties and the topology of the surface. We will explain the 
numerical approaches in detail and discuss various model extensions. 

EU Regional School - Hennig Seminar

Location: AICES Seminar Room 115, 1st floor, Schinkelstr. 2, 52062 Aachen

Prof. Philipp Hennig, Ph.D - Probabilistic Numerics — Uncertainty in Computation

Department of Probabilistic Numerics, Max Planck Institute for Intelligent Systems, Tübingen

Abstract

Data Analysis and Machine learning are prominent topics of contemporary computer science. Their computational complexity is dominated by the solution of non-analytic numerical problems (large-scale linear algebra, optimization, integration, the solution of differential equations). But a converse of sorts is also true: numerical algorithms for these tasks are learning machines! They estimate intractable quantities from observable (computable) quantities. Because they also decide which numbers to compute, these methods can be interpreted as autonomous inference agents. This observation lies at the heart of the emerging area of probabilistic numerical computation, which applies the concepts of probabilistic inference to the design of algorithms, assigning a notion of uncertainty to the result of even deterministic computations. I will outline how this viewpoint is connected to that of classic numerical analysis, then show some concrete examples of algorithms that use the probabilistic formalism to address contemporary algorithmic challenges in large-scale machine learning.