In September, Dr. Rebekka Buse (Statistical Methods & Econometrics at KIT) joined us for a talk entitled "Econometric Methods For Dynamic Networks". Rebekka is a postdoctoral researcher at KIT, Institute for Statistics. She received her PhD from KIT in 2019, during which she collaborated with the Bank for International Settlements in Basel. Previously, she worked at Leibniz University Hannover, Humboldt University Berlin and ENSAE Paris. Her research interests include network measures and time series methods for financial and energy data.

Abstract

After giving an introduction to measuring dynamic networks, we will dive into understanding the European credit risk network and how this was affected by policies and regulations during the sovereign debt crisis. We will explore relevant aspects to consider when measuring networks and further applications in industry and electricity markets.

In July, KCDS Fellows presented their research in:

1. Neural Nets for Solving Economic Models (Lukas Frank, ECON)
2. Data driven model for weather forecasting (Deifilia To, SCC/IMKTRO)
3. Reconstruction of Particle Position and Size in Dispersed Multiphase Flows using Deep Learning and Physics-Based Optimisation (Christian Sax, ISTM/IANM)

Abstracts

1. Neural Nets for Solving Economic Models (Lukas Frank, ECON)

Solving rich economic models globally often requires to solve a high-dimensional functional equation. With classical grid-based methods, the curse of dimensionality limits the number of model features to d ≈ 20. Recent advances leverage the abilities of neural nets to mitigate the curse of dimensionality, bringing more realistic models in reach. However, neural nets pose new challenges such as mediocre accuracy and fragile convergence behavior. I show how to solve high-dimensional economic models with neural nets and how to cure some of the most salient issues.

2. Data driven model for weather forecasting (Deifilia To, SCC/IMKTRO)

Traditional methods for weather forecasting are based on the solution of physical conservation equations that are grounded in theory. In contrast, current machine learning methods learn only through data. Machine learning methods can now create better forecasts than traditional methods - but their success is not well understood. I replicate and study one of the most well-known models, Pangu-Weather, and propose improvements in the architecture that could lead to more efficient training and accurate weather forecasts.

3. Reconstruction of Particle Position and Size in Dispersed Multiphase Flows using Deep Learning and Physics-Based Optimisation (Christian Sax, ISTM/IANM)

Dispersed multiphase flows play an important role in a multitude of environmental and industrial applications, such as spray, mist, cavitation and boiling. A novel diagnostic tool is developed for the investigation of such flows from single camera images. The approach combines deep learning for image segmentation and classification with the optimization of a non-linear functional incorporating a model of the scattering process.

In June, Dr. Anika Rohde (Energy system modeling engineer at TransNet BW) joins us for a talk entitled "Bridging Science and Industry: A Scientist's Path to Electricity Transmission System Operator TransnetBW". 

Abstract

Navigating career paths after a PhD can be challenging. This talk aims to inspire graduates by sharing a personal experience that illustrates the transition from academia to industry. With a background in environmental sciences, Anika Rohde conducted research with the ICON-ART modelling system at the Institute of Meteorology and Climate Research at KIT. After a brief postdoc phase, she transferred to TransnetBW, where she now works in system planning. TransnetBW operates, plans, and expands the electricity transmission grid in Baden-Württemberg. Anika's job focuses on modeling a CO2-neutral energy system for 2050, considering the effects of climate change. This work is crucial for supporting the energy transition and building a sustainable and resilient energy infrastructure.

In May, Dr. Martina Klose (Institute of Meteorology and Climate Research, Department Troposphere Research at KIT) joins us for a talk entitled "Impact of kilometer-scale grid resolutions on Weather and climate modeling: the example of mineral dust".

Abstract:

The contribution of aerosols to the Earth’s changing energy budget is still subject to considerable uncertainty, not least due to limited process-level understanding. Mineral dust from dry soils is the dominant aerosol type in terms of global mass and it is also amongst the most important substrates for cloud formation through heterogeneous ice nucleation. Dust therefore has important impacts on climate, but also on air quality and health, road and air traffic, economy, and not least photovoltaic power output in many areas of the world including Europe. Dust emission is a threshold process that depends non-linearly upon surface wind intensity, which means that the accuracy at which models represent surface winds is key to estimate dust emissions. In this talk, I will present how new global storm-resolving models at single-digit kilometer grid resolutions offer unprecedented insights into the contribution of smaller-scale, but intense wind systems and related dust storms, which cannot be represented at coarser model resolutions.

The KIT Graduate School Computational and Data Science (KCDS) at KIT Center MathSEE proudly presents: KCDS Talks, a monthly series of short lectures from basic knowledge to trending topics in computational and data science.

In April, Jun.-Prof. Dr. Alina Roitberg (University of Stuttgart - Institute for Artificial Intelligence and KIT Alumna) joins us for a talk entitled "From Representation to Action and Intention: Learning to Perceive Humans in a Changing World".

Abstract

This talk will explore recent advances in video-based human activity recognition aimed at creating adaptable, resource-efficient, and uncertainty-aware models for assisting humans in everyday situations. Topics covered will include: (1) an overview state-of-the-art methods and  public datasets for human activity recognition in different assistive scenarios  (2) the importance of adaptability in human observation systems to cater to new situations (environments, appearances, behaviours) as well as strategies for addressing such open world tasks, and (3) incorporating uncertainty-aware approaches, vital for robust and safe decision-making. The talk will conclude with a discussion of future research directions and the potential applications of these models, such as technology for elderly assistance or medical diagnostics.

Bio

Alina Roitberg is a Tenure-Track Juniorprofessor at the University of Stuttgart, leading the newly established Intelligent Sensing and Perception Group at the Institute for AI, University of Stuttgart. She is also a Faculty member of the International Max Planck Research School for Intelligent Systems (IMPRS-IS). Before joining the University of Stuttgart, she was a postdoctoral researcher at KIT and a Visiting Researcher at Johannes Kepler University Linz. She received her PhD from KIT in 2021, during which she completed a research internship at Facebook Zurich. For her doctoral work, she was recognized with multiple awards, including the IEEE ITSS Best Dissertation Award and the Helmholtz Doctoral Prize. Her research interests include computer vision, human activity recognition, domain adaptation, open set recognition, as well as resource- and data-efficient learning.

In March 2024, Dr. Annika Oertel (KIT) joined us for a talk entitled "Quantifying parametric uncertainty in numerical weather prediction models". 

In the February 2024 edition of KCDS Talks, KCDS Fellow Orhan Delil Tanrıkulu presented his doctoral project entitled "Exploring the Potential of machine learning methods for improving operational hydrological forecasting and prediction (EPOforHydro)".

KCDS Talks kicked off the new year with a talk by guest speaker and KIT alumn Prof. Dr. Markus A. Stricker of ICAMS at Ruhr-University Bochum, who presented "Materials Informatics - Appreciation of data and algorithms".

Slides are available on the KCDS Talks course group on ILIAS.

In November, KCDS Fellows Siyu Li and Georgios Evangelopoulos presented their previous work and doctoral projects:

• "Development of a machine learning model for hail prediction in Germany" + "Advancing Subseasonal Predictions at Reduced Computational Effort - Identifying source regions and pathways of tropical forecast errors towards the extratropics", Siyu Li (IMK-TRO)
• "Optimization and acceleration of aerosol dynamics and chemistry sub-processes in weather and climate models", Georgios Evangelopoulos (SCC)

In October 2023, KCDS Fellows Felix Rein, Nicholas Popovic and Manoj Mangipudi presented their previous scientific work and their doctoral projects:

• Felix Rein (IMK-TRO) - "Evaluation of precipitation forecasts in tropical West Africa: Numerical Weather Prediction vs Climatology"
• Nicholas Popovic (AIFB) - "Representation Learning for Relation Extraction"
• Venkata Surya Siva Manoj Mangipudi (SCC) - "Machine Learning methods to develop predictive models for estimation of exhaust gas properties from internal combustion engines during cold starts"

In September 2023, PD Dr. Andreas Ott (KIT) joined us for a talk on "Applications of Topological Data Analysis in the Life Sciences".

This issue of KCDS Talks was also a first birthday party for KCDS! The agenda included presentations of the following KCDS Fellows on their previous scientific work and doctoral projects:

• Jaroslav Borodavka (SCC): "Effective data-driven training strategies for complex multiscale stochastic differential equations"
• Felipe Donoso (IBPT): "Mathematical modeling and simulation of terahertz waves, photons and sub-pstransients towards terahertz tomography at accelerators and for medical applications"
• Miriam Simm (IMK): "Post-simulation diagnostics of microphysical process rates from a climate and weather model with AI"
• Anna Reiffer (IFV): “Generation and Scheduling of Activities for Travel Demand Models to Account for Telecommuting Behaviour”

In May 2023, Dr. John A. Warwicker (KIT) joined us for his talk on "A Unified Framework For Clustering And Regression Problems Via Mixed-integer Linear Programming".

In April 2023, Dr. Johannes Bracher (KIT) joined us for his talk on "Forecasts in Epidemiology".

In March 2023, Dr. Cihan Ates (KIT) joined us for a talk entitled "How the brain learns and why adaptive models matter".

Slides, literature and a video recording of Cihan's talk are available in the KCDS Talks course group on ILIAS.

In February 2023 we had the premiere of our lecture series KCDS Talks!

PD Dr.-Ing. Uwe Ehret (KIT) held the first lecture on "Basics of Information Theory".

Slides, literature and a video recording of Uwe's talk are available in the KCDS Talks course group on ILIAS.