Here you can find our seminars, respectively the regular modelling seminar. As always, registration takes place via Moodle, where you can also see the responsible lecturers for each semester. If you have any questions, we are happy to help you!
Modelling seminar
In the modelling seminar, you will learn about the various application areas of mathematics and model them using simple examples.
Typ: seminar (2 SWS)
Tonus: every term
Language: German
Prior knowledge: Basic mathematical knowledge is assumed.
Module: SpdWA, SpdWB, K-MAT1
Examination: In order to receive the ECTS, regular participation in the seminar, a presentation and a written paper (approx. 10 pages) are required.
In the modelling seminar, questions in physics, biology, geology, economics, etc. are discussed and mathematics is looked at from a completely new angle: Mathematics is included in many areas of our lives!
In the modelling seminar, participants discuss and solve problems and phenomena in small groups with the help of mathematics, e.g. in the field of medicine the effects and effectiveness of vaccination strategies. The models obtained in the seminar are calibrated using freely available data. They allow an outlook on (possible) future developments and also the influence of parameters on the problem variables. This seminar is equally open to mathematicians, economists, teachers and combined bachelor students.
Literature:
- M. Lehn, Wie halte ich einen Seminarvortrag, (pdf-file)
- S.P. Jones, How to write a good research paper and give a good research talk
- I. Parberry, How to present a paper in theoretical computer science, Bulletin of the EATCS,(37), 1989.
- S.P. Jones, J. Launchbury, J. Hughes, How to give a good research talk, SIGPLAN Notices 28(11), November 1993.
- G. Aiglstorfer, A short guide for student talks and papers, TU Munich, 2004.
- H. Kraft, Das Verfassen und Präsentieren wissenschaftlicher Arbeiten, TU Munich, 2006.
Seminar Uncertainty in Deep Learning
In this course you will study selected papers on the topic of uncertainty quantification (UQ) in deep learning (DL). UQ is particularly important when DL shall be used in safety-critical applications, such as automated driving, medical diagnosis, medical imaging, etc.
Type: Seminar
Tonus: Winter-semester
Language: English / (German)
Prior Knowledge: ---
Examination: You are expected to present a corresponding paper in 30 minutes (+ ~10 minutes discussion) and to attend the seminar regularly.
In this course we study selected papers on the topic of uncertainty quantification (UQ) in deep learning (DL). UQ is particularly important when DL shall be used in safety-critical applications, such as automated driving, medical diagnosis, medical imaging, etc. Furthermore, UQ as multiple applications, ranging from the detection of malfunction, detection of label errors in the data, over active learning to human machine collaboration. The goal of this seminar is to get familiar with recent methodological developments in UQ for DL. For your presentation, you are expected to
- Get familiar with the respective paper and related literature on your own;
- Pre-process the paper with emphasis on didactics such that other students can understand the idea, background and method;
- Present the experiments and the conclusion obtained from them.
