Master-Seminar - Deep Learning in Physics (IN2107, IN0014)

Prof. Dr. Nils ThuereyPatrick Schnell, Björn List

Time & place

Every Monday, 12:00-14:00 in room: MI 02.13.010

Begin:

 

Monday, April. 15., 2024

 

 

Content

Using deep learning methods for physical problems is a very quickly developing area of research. The research group of Prof. Thuerey has studied learning-based methods for Navier-Stokes problems and fluid flow applications in recent years, examples of which include learning latent-spaces for physical predictions, generative adversarial networks with temporal coherence, and the inference of Reynolds-averaged Navier-Stokes flows around airfoils. Beyond these physics-based deep learning works of the Thuerey group, this seminar will give an overview of recent developments in the field.

In this course, students will autonomously investigate recent research about machine learning techniques in the physical simulation area. Independent investigation for further reading, critical analysis, and evaluation of the topic are required.

Paper Selection

The paper list for this seminar can be found below. Please send us an email with your 5 ranked preferences by Sunday, April 7th. 

We will assign papers based on these preferences, while remaining papers will be distributed randomly. The paper-matching results will be made available before the Kick-off meeting.

Requirements

Report

When:
  • semi-final version is due one week before your talk (Monday by 23:59)
  • Send your final report within two weeks after your talk (Monday by 23:59).
Style:
Guidlines:
  • You can begin with writing a summary of the work you present as a start point; but, it would be better if you focus more on your own research rather than just finishing with the summary of the paper. We, including you, are not interested in revisiting the work done before; it is more meaningful if you make an effort to put your own reasoning about the work, such as pros and cons, limitation, possible future work, your own ideas for the issues, etc.

Slides

When:
  • Send semi-final slides at least one week before your presentation, together with the semi-final report. Otherwise the presentation will be cancelled. Please also make an appointment with your advisor when you send your slides. There is an mandatory discussion with your advisor in the week before your presentation. Your advisor will give your feedback on your slides. 
  • Send final slides within two weeks after your presentation to us (Monday by 23:59).
Style:
  • Any slide style you like, prepare slides as PDF file.
Guidelines:
  • Ensure readability (colors, images and font size).
  • Avoid using too much text.
  • Highly encouraged to do some paper-related experiments and show some results in the presentation.

Presentation

  • Present your topic in English.
  • You have 25 minutes for presentation and 10 minutes for questions and discussion.
  • Please actively participate in the discussion for other presentations.
  • Please test your setup (laptop/connection to projector) before giving your presentation!

Attendance

  • Missing one session is allowed, if you let us know in advance and write a short summary of the papers (ca. 1 page) in your own words.
  • Missing another session means failing the seminar (special rules for severe issues as appropriate).

Paper list

Nb. Title      
1 Physics Informed Deep Learning: Data-driven Solutions of Nonlinear Partial Differential Equations      
2 DeepONet: Learning nonlinear operators for identifying differential equations based on the universal approximation theorem of operators      
3 Fourier Neural Operator for Parametric Partial Differential Equations      
4 Learning to control PDEs with differentiable physics      
5 Do Differentiable Simulators Give Better Policy Gradients?      
6 Accelerated Policy Learning With Parallel Differentiable Simulation      
7 On the difficulty of learning chaotic dynamics with RNNs      
8 The curse of unrolling      
9 low variance gradient optimization in unrolled computation graphs with es single      
10 Automatic Symmetry Discovery with Lie Algebra Convolutional Network      
11 Lagrangian Fluid Simulation with Continuous Convolutions      
12 Message Passing Neural PDE Solvers      
13 PDE-Refiner: Achieving Accurate Long Rollouts with Neural PDE Solvers      
14 Convolutional Neural Operators for robust and accurate learning of PDEs      
15 Learning skillful medium-range global weather forecasting      
16 Machine learning–accelerated computational fluid dynamics      
17 Deep learning for universal linear embeddings of nonlinear dynamics      
18 SINDy-PI: a robust algorithm for parallel implicit sparse identification of nonlinear dynamics      
19 Learning data-driven discretizations for partial differential equations      
20

Learning Physics Constrained Dynamics Using Autoencoders

     

 

You can access the papers through TUM library's eAccess.

Preliminary Schedule

Date Paper Number Presenter  Advisor
15.04.24 Kick-off lecture    
22.04.24 no seminar    
29.04.24 no seminar    
06.05.24 1 Medrano Navarro Patrick
06.05.24 2 Valiullin Patrick
06.05.24 11 Timár Bjoern
13.05.24 14 Sutar Bjoern
13.05.24 20 Pavlova Bjoern
20.05.24 holiday - no seminar    
27.05.24 3 Pfister Patrick
27.05.24 4 Tikhomirov Patrick
03.06.24 6 Meyering Patrick
03.06.24 16 Metscher Bjoern
03.06.24 8 Yücel Patrick
10.06.24 7 Antrich Patrick
10.06.24 15 Redinger Bjoern
17.06.24 9 Köse Patrick
17.06.24 13 Alizada Bjoern
17.06.24 18 Nguyen Bjoern
       

Resources

Contact any time you have questions related to the seminar or your paper!

Kickoff Slides