APEBench Talk @ Pasteur Labs Journal Club

Pasteur Labs (https://simulation.science/) invited me to their Journal Club to talk about my recent NeurIPS paper, which is on benchmarking autoregressive neural emulators for predicting spatiotemporal phenomena. Here are the slides: https://github.com/tum-pbs/apebench-p...

APEBench paper: https://arxiv.org/abs/2411.00180
APEBench project page: https://tum-pbs.github.io/apebench-pa...
APEBench code: https://github.com/tum-pbs/apebench

APEBench is a project I worked on as part of my PhD with Prof. Thuerey at TUM. If you like APEBench, I am sure you will enjoy our other research on deep learning in physics simulations: https://ge.in.tum.de/

-------

📝 : Check out the GitHub Repository of the channel, where I upload all the handwritten notes and source-code files: https://github.com/Ceyron/machine-lea...

📢 : Follow me on LinkedIn or Twitter for updates on the channel and other cool Machine Learning & Simulation stuff:   / felix-koehler   and   / felix_m_koehler  

---

Timestamps:

00:00 Introduction
00:30 Spatiotemporal Simulation (with classical numerical methods)
02:44 What is Autoregressive Emulation?
03:38 How to train emulators using synthetic data?
06:30 Abstract Pipeline for PDE Datasets & Benchmarks
07:43 Problems in the Pipeline
09:36 What APEBench does differently
10:52 Key Contribution: A fast reference solver (ETDRK)
12:14 PDE Identifiers (difficulty-based interface)
16:05 Possible axes of investigation with APEBench
17:15 Sample Study: Advection Emulation
21:27 Use cases of APEBench
22:49 APEBench summary
25:33 Conclusion & Outro