Exchange in Scientific Computing (ESC)
A distinguished series of lectures on topics from the field of scientific computing with social and scientific relevance.
Atomistic Simulations of Complex Systems with High-Dimensional Neural Network Potentials
Prof. Dr. Jörg Behler: Lehrstuhl für Theoretische Chemie II, Ruhr-Universität Bochum
and
Research Center Chemical Sciences and Sustainability, Research Alliance Ruhr
HS C - Technik
6020 Innsbruck
17:15 - 18:30
Abstract
Machine Learning Potentials (MLPs) have revolutionized the field of atomistic simulations, since they allow to combine the accuracy of electronic structure calculations with the efficiency of empirical potentials thus enabling the investigation of large systems on long time scales. When selecting a particular type of MLP, the physical interactions in the system of interest have to be taken into account to ensure a reliable representation of the multidimensional potential energy surface. In this talk, an overview of four generations of high-dimensional neural network potentials will be given, from efficient local potentials to MLPs explicitly considering long-range interactions and non-local charge transfer. Typical applications for systems of increasing complexity will be discussed with a focus on simulations of solid-liquid interfaces.
Jörg Behler studied chemistry at the University of Dortmund from 1995 to 2000, completing his studies with an external diploma thesis at the University of Reading in Great Britain. He then completed his doctoral dissertation at the Fritz Haber Institute of the Max Planck Society in Berlin. After receiving his doctorate in 2004, he initially conducted research there and, from 2006, as a postdoctoral researcher at ETH Zurich. In 2007, he moved to the Chair of Theoretical Chemistry at Ruhr University Bochum, where he led an independent junior research group from 2008 to 2017. This group was funded by a Liebig Fellowship as well as the Emmy Noether Programme and the Heisenberg Programme of the German Research Foundation (DFG). He completed his habilitation in 2014. In 2017, he was appointed professor at the Georg-August University of Göttingen, where he held the Chair of Theoretical Chemistry, until 2022 when he was appointed to the newly established Chair of Theoretical Chemistry II at the Faculty of Chemistry and Biochemistry of the Ruhr University and a research professorship at the Research Center "Chemical Sciences and Sustainability" of the Research Alliance Ruhr.
HPC & AI - Competition or Collaboration
Prof. Dr. Erwin Laure, Max Planck Computing and Data Facility (MPCDF), Germany
HSB 6 - Technik
6020 Innsbruck
17:15 - 19:00
Abstract
Since Generative AI has become mainstream through Large Language Models like e.g. employed in ChatGPT or DeepSeek, AI is increasingly considered as a potential tool in scientific workflows. While classical AI is in mainstream use in image based research (e.g. for analyzing brain scans) for many years, other domains are still in the explorative phase. But this is changing at an enormous speed as e.g. exemplified by the recent announcement of ECMWF to use AI in their weather forecast. Yet, how far AI can replace classical simulations, is still subject to ongoing debates.
At the same time, AI has a profound impact on HPC hardware industry. Double precision, typically employed in scientific simulations, is not needed for AI and chip manufacturers start to reduce double precision capabilities in favour of low precision units. This is not surprising, given an AI market that is several orders of magnitudes large than the HPC one.
In this talk we review some of the impact AI has made in scientific computing, using examples from practical AI use within the Max Planck Society. We also review the impact, AI has on hardware industry and how this affects classical scientific computing. Whatever the future will bring: AI has come to stay and while it is a competition to classical HPC in some respect, those being able to effectively exploit AI capabilities will likely have a competitive advantage.
Erwin Laure is the Director of the Max Planck Computing and Data Facility (MPCDF) of the MPG in Garching, Germany and Honorary Professor at the Technical University Munich. Before joining MPG he was Professor for High Performance Computing at KTH Stockholm and Director of the PDC Center for High Performance Computing there. He holds a PhD from the University of Vienna and has more than 25 years experiences in High Performance Computing, was a member of EuroHPC Infrastructure Advisory group and involved in major European Exascale projects (e.g. the BioExcel Centre of Excellence for Biomolecular Simulations). His research interests include programming environments, languages, compilers and runtime systems for parallel and distributed computing, with a focus on exascale computing.
Fast Parallel Algorithms for Large-Scale Computational Problems
Prof. Dr. Ulrich Rüde, Friedrich-Alexander-Universität (FAU)
HS - Technik
6020 Innsbruck
17:00 - 19:00
Free drinks after the talk for all!
Abstract
Modern high-performance computers enable simulations at scales that were unthinkable only a few decades ago. At the same time, the mathematical methods and algorithms that underpin these computations have become an even more critical factor to achieve efficiency and accuracy. This talk will discuss recent developments in fast numerical methods and iterative solvers for large-scale scientific computing. Emphasis will be placed on the design and analysis of parallel algorithms, their implementation on current and emerging computing architectures, and their role in advancing simulation-based research in science and engineering.
Ulrich Rüde studied mathematics and computer science, earning a Master’s degree from the Florida State University and a Ph.D. from Technische Universität München. From 1998 to 2025 he held the Chair of System Simulation at Friedrich-Alexander-Universität Erlangen-Nürnberg, where he built an internationally visible research program in Computational Science and Engineering (CSE). He also led the Algo-COOP team at CERFACS in Toulouse and was a visiting professor at the University of Colorado, the National University of Singapore, and Université de Rouen Normandie. Since 2024 he has been a Senior Researcher at VSB – Technical University of Ostrava. His research focuses on high-performance computing, algorithms, and scalable numerical methods for large-scale simulations across scientific and engineering domains.
High Performance Models, Methods and Computing - from Basics to Applications in Medicine, Natural Sciences and Engineering
Prof. Dr.-Ing. Wolfgang A. Wall, TUM
HSB 1 - Technik
6020 Innsbruck
17:00 - 19:00
Abstract
Scientific Computing is meanwhile widely acknowledged as the third pillar of scientific discovery and also an essential every day tool in many industries. However, besides impressive growth rates, both on the academic and the commercial side, the full potential of it is not nearly revealed yet. A key to unchain the full potential is to combine all the relevant building blocks at the best possible level - the art of modeling, the development of computational methods as well as the appropriate implementation ready for the best HPC platforms. This talk will try to showcase that such a combination can create game changers in different fields in Science and Engineering and can even lead to a paradigm-shift in health care.
Wolfgang A. Wall is full Professor and founding Director of the Institute for Computational Mechanics at the Technical University of Munich, Germany. Born near Salzburg (Austria), he studied at the University of Innsbruck and received his PhD from the University of Stuttgart. Among others, he acted as founding director of the Munich School of Engineering and is co-founder of the companies AdCo EngineeringGW and Ebenbuild GmbH. Wolfgang A. Wall has received several esteemed awards, is a recent recipient of an ERC Advanced Grant and serves on a large number of prestigious boards. He currently also serves as Rector of the International Centre for Mechanical Sciences (CISM) in Udine (Italy) and is member of the Austrian Academy of Sciences as well as of the Bavarian Academy of Sciences and Humanities.
Powering the Future: Tackling Energy Challenges in Supercomputing and AI
Prof. Dr. Dieter Kranzlmüller, LMU Munich
AULA
6020 Innsbruck
18:00 - 20:00
Abstract
Recent performance gains in supercomputing are being outpaced by soaring power consumption, especially in leadership-class supercomputers. AI infrastructures face similar challenges, with large language models driving energy demands to critical levels. In response, data centers are expanding their power and cooling capacities, adding further to their global environmental footprint. The need for improved energy efficiency in computing has never been more urgent. Have we already reached a critical tipping point, or can innovation drive a more sustainable future? This talk explores the latest advances and examines strategies for balancing the growing energy demands of advanced computing with environmental responsibility.
Slides
The speaker generously provides the slides for download.
Dieter Kranzlmüller is full professor for Computer Science at the Ludwig-Maximilians-Universität (LMU) in Munich, and chairman of the board of directors of the Leibniz Supercomputing Centre (LRZ), an institute of the Bavarian Academy of Sciences and Humanities. He is a board member of the national Gauss Center for Supercomputing (GCS), on the senate of the National Research Data Infrastructure (NFDI), and the founding convent of the Interdisciplinary Transformation University Austria (IT:U). His research interests concern data center management and operation, high-performance computing and AI, quantum computing, and future computing technologies.