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

LecturerKiwon Um, Philipp Holl, and Nils Thuerey
StudiesMaster Informatics
TimeMondays, 12:00-14:00
PlaceSeminarraum MI 02.13.010
BeginApril 29., 2019

Contents

Deep learning 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 studies, 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 physics. Independent investigation for further reading, critical analysis, and evaluation of the topic are required.

Requirements

Participants are required to first read the assigned paper and start writing a report. This will help you prepare for your presentation.

Attendance
  • It is only allowed to miss a single time-slot. Missing a second one means failing the seminar. If you have to miss any, please let us know in advance.
Report
  • Your final report (4 pages max. excluding references in the ACM SIGGRAPH TOG format (acmtog) - you can download the precompiled latex template) should be prepared and sent no later than two weeks before your talk, i.e., by 23:59 on Monday.
  • Guideline: 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.
Presentation (slides)
  • You will present your topic in English, and the talk should last 30 minutes. After that, a discussion session for ca. 10 minutes will follow.
  • The slides should be structured according to your presentation. You can use any layout or template you like.
  • Plagiarism is important; please do not simply copy the original authors' slides. You can certainly refer to them.
  • The semi-final slides (PDF) should be sent one week before the talk; otherwise, the talk will be canceled.
  • We strongly encourage you to finalize the semi-final version as far as possible. We will take a look at the version and give feedback. You can revise your slides until your presentation.
  • Be ready in advance. Please make sure your presentation (including videos/demos) runs on your laptop even without Wi-Fi. We can provide you with an adapter for the projector and a clicker. If you need anything else (e.g., speakers or a laptop), please contact us in advance.
 

Preliminary Schedule

08 Mar 2019Deregistration due
12 Apr 2019Send three preferred topics
19 Apr 2019Assign topics
29 Apr 2019Introduction lecture 
27 May 2019First talk

Papers

NoDate Presenter Paper
01Canceled 2015, Ladicky et al., Data-driven Fluid Simulations using Regression Forests, TOG
0227 May Jonas Maximilian Weigand 2017, Tompson et al., Accelerating Eulerian Fluid Simulation With Convolutional Networks, PMLR
0303 Jun Lennart Pauli 2016, Li et al., To Fall Or Not To Fall: A Visual Approach to Physical Stability Prediction, arXiv.org
04Canceled 2017, Watters et al., Visual Interaction Networks: Learning a Physics Simulator from Video, NIPS 2017
0517 Jun Chinmay Prabhakar 2018, Schenck and Fox, SPNets: Differentiable Fluid Dynamics for Deep Neural Networks, PMLR
0617 Jun Dominik Mehringer 2018, Luo et al., DeepWarp: DNN-based Nonlinear Deformation, arXiv.org
0724 Jun Dávid Endrédi 2018, Xie et al., tempoGAN: A Temporally Coherent, Volumetric GAN for Super-resolution Fluid Flow, TOG
0824 Jun Taha Emre 2018, Kim et al., Deep Fluids: A Generative Network for Parameterized Fluid Simulations, arXiv.org
0901 Jul Berkay Alp Cakal 2018, Erhardt et al., Unsupervised Intuitive Physics from Visual Observations, arXiv.org
1001 Jul Johannes Kroll 2018, Baque et al., Geodesic Convolutional Shape Optimization, arXiv.org
11Canceled 2018, Pathak et al., Model-Free Prediction of Large Spatiotemporally Chaotic Systems from Data: A Reservoir Computing Approach, PRL
12Canceled 2018, Beck et al., Deep Neural Networks for Data-Driven Turbulence Models, arXiv.org
1315 Jul Linus Seidler 2017, Holden et al., Phase-functioned Neural Networks for Character Control, TOG
1415 Jul Malek Souissi 2018, Long et al., PDE-Net: Learning PDEs from Data, PMLR
1522 Jul Simon Langrieger 2018, Mrowca et al., Flexible Neural Representation for Physics Prediction, NIPS
1622 Jul Vindhya Singh 2018, Raissi and Karniadakis, Hidden Physics Models: Machine Learning of Nonlinear Partial Differential Equations, JCP

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