2021 - KoMSO Academy

Combining model- and data-based approaches for industrial problems, September 14-16, 2021

AI industrial problems

The Center for Industrial Mathematics (ZeTeM) and the Committee for Mathematical Modeling, Simulation, and Optimization (KoMSO) are happy to host the first KoMSO Academy: Combining model- and data-based approaches for industrial problems in 2021.

The KoMSO Academy combines a Challenge Workshop with well-known speakers from industry and academia and a Training course with hands-on session. The focus of the Challenge Workshop will be on energy and mobility.

The organization of the KoMSO Academy is dependent on the pandemic situation. We will held a hybrid event with limited on-site access and online participation.Please note that the 3G rule currently applies to all events in Bremen.

On-site registration is closed!

Keynote Lecture

Wil Schilders, Eindhoven University of Technology

These are exciting times for mathematicians working in the area of Computational Science and Engineering (CSE). The hardware on which we run our simulations has changed dramatically, data science has come into the picture, and artificial intelligence (AI), especially machine learning (ML), has become omnipresent. Mathematicians must adapt to this new situation and come up with entirely new methods and use out-of-the-box thinking.

This holds especially true for the field of machine learning, where more and more criticism is heard: we do not understand why methods work or do not work, it is unclear which features are used to take decisions, the topology of networks is often guesswork, the concepts of time, space and causality are not used, and many more issues. On the other hand, researchers within the CSE area have realized that machine learning and neural networks could aid in building more accurate models, and in this way contribute to, for example, digital twinning where extremely accurate models and efficient simulations are highly necessary. The USA is a front-runner in the field of combining CSE and AI/ML methods, but Europe is in an excellent position to contribute significantly to this exciting new area.

In the lecture, we will start with some opening thoughts and a brief history of AI and ML, followed by examples of how we can make bridges between CSE and AI to benefit advanced simulations, especially those necessary in an industrial context.


Training and hands-on session

The hands on sessions will be given by Dr. Daniel Otero Baguer. All sessions will use Python as programming language and the PyTorch framework.

Topics
  • Introduction to training neural networks with PyTorch
  • Model-based classic approaches, e.g., ISTA.
  • Introduction to data-based methods, e.g., LISTA
  • Neural Networks for trivial ill-posed inverse problems and fully data-based methods
  • Combining model and data based methods: learned post-processing and learned gradient descent
  • Deep Image Prior and mathematical aspects
  • Applications on Computed Tomography (CT)
Requirements
  • Some experience programming in Python (if not you can still follow the sessions but it will be much harder).
  • Google Chrome web browser installed.
  • A google account, we are going to use Google Colab: https://colab.research.google.com/

Challenge Workshop

Challenge Workshops are adressing novel business and technology trends at an early stage of development. They bring together leading experts from industry and academia for a discussion of the present state of the art, potentials and risks as well as future developments. They are open to short contributions by the participants aiming for feedback for their specific problems. The results of the workshops are summarized in a white paper aiming to support policymakers, researchers, entrepreneurs and the public to debate and shape the future of research and innovation in the respective field.

This workshop will focus on the combination of model- and data-based approaches for industrial problems in the field of mobility and energy.

Confirmed speakers

Ralph Grothmann, Siemens AG, Bremen
Julian Bergmann, Chris Michel Peters, EWE AG DataLab, Oldenburg
Björn Weißhaupt, Deutsche Bahn AG, Frankfurt
Ralf Korn, TU Kaiserslautern
Wil Schilders, Eindhoven University of Technology
Hanno Gottschalk, University of Wuppertal

 


Place and Agenda

The training and hands-on session will take place at the University of Bremen, building MZH, room 1470. Online participation is possible and the dial-in data will be provided by E-mail.

Please find the agenda below.

Place and Agenda

The Challenge Workshop will take place at the Haus der Wissenschaft, Sandstraße 4/5. Online participation is possible and the dial-in data will be provided by E-mail.

Please find the agenda below.

Organizers & Registration

Organizers:
Prof. Dr. Dr. h.c. Peter Maaß, Universiy of Bremen
Dr. Daniel Otero Baguer, University of Bremen
Dr. Lena Hauberg-Lotte, University of Bremen
Dr. Johannes Herold, University of Heidelberg, KoMSO

Registration:
Date: September 14-16, 2021
Place: Bremen or virtual
Registration fee for industry: Training and hands-on session 500 €/ Workshop 300 €
Registration fee for academia: 150 €

For registration please send an e-mail to the local organization comittee.

komso-academyprotect me ?!math.uni-bremenprotect me ?!.de

KI Gehirn