Highdimensional data are implicitly obtained in classification and regression tasks by certain supervised learning algorithms like e.g. Support Vector Machine through Mercer's Theorem. An example where data is explicitly presented in high dimension is given by hyperspectral data. Due to the "curse of dimensionality", dimension-reduction methods become important. Since data are often non-linear, manifold learning techniques are of interest.
The 3-day workshop "Topological Data Analysis and Manifold Learning using Ultrametrics" aims to introduce methods inspired by topology and manifolds for the analysis of high-dimensional data, as well as to practically incorporate some novel ideas from ultrametric analysis in order to obtain faster algorithms. The idea is to test these methods on datasets of interest by the participants and to aim at collaboratively producing novel experimental results, at least in the aftermath of this workshop.
This workshop is brought to you in cooperation of graduate schools KCDS and GRACE. Everyone interested is welcome to join! - For graduate school members, successful participation will be credited with 2 CP.

Highdimensional data are implicitly obtained in classification and regression tasks by certain supervised learning algorithms like e.g. Support Vector Machine through Mercer's Theorem. An example where data is explicitly presented in high dimension is given by hyperspectral data. Due to the "curse of dimensionality", dimension-reduction methods become important. Since data are often non-linear, manifold learning techniques are of interest.
The 3-day workshop "Topological Data Analysis and Manifold Learning using Ultrametrics" aims to introduce methods inspired by topology and manifolds for the analysis of high-dimensional data, as well as to practically incorporate some novel ideas from ultrametric analysis in order to obtain faster algorithms. The idea is to test these methods on datasets of interest by the participants and to aim at collaboratively producing novel experimental results, at least in the aftermath of this workshop.
This workshop is brought to you in cooperation of graduate schools KCDS and GRACE. Everyone interested is welcome to join! - For graduate school members, successful participation will be credited with 2 CP.

Highdimensional data are implicitly obtained in classification and regression tasks by certain supervised learning algorithms like e.g. Support Vector Machine through Mercer's Theorem. An example where data is explicitly presented in high dimension is given by hyperspectral data. Due to the "curse of dimensionality", dimension-reduction methods become important. Since data are often non-linear, manifold learning techniques are of interest.
The 3-day workshop "Topological Data Analysis and Manifold Learning using Ultrametrics" aims to introduce methods inspired by topology and manifolds for the analysis of high-dimensional data, as well as to practically incorporate some novel ideas from ultrametric analysis in order to obtain faster algorithms. The idea is to test these methods on datasets of interest by the participants and to aim at collaboratively producing novel experimental results, at least in the aftermath of this workshop.
This workshop is brought to you in cooperation of graduate schools KCDS and GRACE. Everyone interested is welcome to join! - For graduate school members, successful participation will be credited with 2 CP.

2-day on-site workshop on November 4+5, 2025 on KIT Campus South
 
As a scientist, presenting your findings is essential, whether in a group seminar, at a conference, or during your Ph.D. defense. However, effective science presentation is more than just sharing data.

The topics covered by this workshop are:
Planning and structure Memorable start and ending Speech, voice, and body language Create meaningful graphics Slide composition Difficult situations  
The training is a workshop, meaning that the participants apply the learned material to their presentations right there. Therefore, everybody should bring presentations (old or upcoming ones) to work with.

Speaking the same language is a great advantage for tailoring presentations effectively. Another important point: Most scientists talk about their topic with shining eyes. Often, that excitement is gone the moment they stand in front of a crowd. To tap into this enthusiasm during your talk is particularly important. For that enthusiasm to reappear, you need to feel safe with what you are doing in your presentation.
 
Brought to you by Institute of Measurement and Control Systems (MRT) and KCDS.

2-day on-site workshop on November 4+5, 2025 on KIT Campus South
 
As a scientist, presenting your findings is essential, whether in a group seminar, at a conference, or during your Ph.D. defense. However, effective science presentation is more than just sharing data.

The topics covered by this workshop are:
Planning and structure Memorable start and ending Speech, voice, and body language Create meaningful graphics Slide composition Difficult situations  
The training is a workshop, meaning that the participants apply the learned material to their presentations right there. Therefore, everybody should bring presentations (old or upcoming ones) to work with.

Speaking the same language is a great advantage for tailoring presentations effectively. Another important point: Most scientists talk about their topic with shining eyes. Often, that excitement is gone the moment they stand in front of a crowd. To tap into this enthusiasm during your talk is particularly important. For that enthusiasm to reappear, you need to feel safe with what you are doing in your presentation.
 
Brought to you by Institute of Measurement and Control Systems (MRT) and KCDS.

Join us for a workshop featuring invited talks by:
 
Sebastian Engelke (University of Geneva)
Sebastian Engelke works on extreme value theory, graphical models, and statistical climate science. His group also focuses on extrapolation methods in machine learning and weather forecasting with AI.
 
Linda Mhalla (École Polytechnique Fédérale de Lausanne)
Linda Mhalla's research focuses on extreme value theory, quantitative risk modelling, and causal inference, with applications in environmental and financial contexts. In her presentation, she will discuss causal discovery in multivariate extremes, applied to environmental data
 
Gabriele Messori (Uppsala University)
Gabriele Messori is an atmospheric physicist whose research focuses on climate extremes, including windstorms, temperature extremes, and heavy precipitation. His work adopts an interdisciplinary approach, examining the physical drivers, predictability, and impacts of these events within the Earth System.
 
Petra Friederichs (University of Bonn)
Petra Friederichs focuses on the statistical modelling and forecasting of weather and climate extremes, including post-processing of forecasts, the development of targeted verification approaches, and the detection and attribution of human activities on the evolution and characteristics of extreme events in the future climate.
 
KCDS members, as well as researchers from KIT and other universities/research centers, are warmly welcome. Participation is free of charge. Please note that KCDS does not cover travel or accommodation expenses. Register by 31 Oct 2025.
 
Organizers: Lisa Leimenstoll and Tobias Biegert
 
Register here: https://indico.kit.edu/event/5224/