🎉 Great news: Our Machine Learning and Physical Sciences workshop will be back again this year! 🎉
Keep an eye out for updates on deadlines etc, we will be updating the website soon
ml4physicalsciences.github.io
#ML4PS2025
Implying there was a point before
Congrats and welcome!!
Are you using the API or Console (console.anthropic.com) to get more Claude? (Happy to pass along any feedback...)
📣 Hiring! I am looking for PhD/postdoc candidates to work on foundation models for science at @ULiege, with a special focus on weather and climate systems. 🌏 Three positions are open around deep learning, physics-informed FMs and inverse problems with FMs.
Yah agreed! Opportunity cost aside, I think a generally nice time as a postdoc, then going to industry seems good from a happiness perspective. It seems likely that regret here mostly reflects a less-than-ideal experience post-PhD (by default for systemic reasons).
(If true) I think that's a very specific case, I think industry employers generally don't care or might even weakly prefer fewer years post-PhD
I remain a contrarian on this!
Got a photo with a nice branch of my academic family tree: @awehwe.bsky.social @glouppe.bsky.social @smsharma.bsky.social @lukasheinrich.com
Nicole Hartman
@annalenakofler.bsky.social
The papers and posters for our Machine Learning and Physical Sciences workshop at #NeurIPS2024 are online #ml4ps2024. Come check it out on Sunday
ml4physicalsciences.github.io/2024/
Poster 4.15, talk 2.30! (Is the schedule shown differently somewhere?)
Looking forward to catching up with friends and colleagues at NeurIPS next week, as well as meeting new folks -- feel free to reach out ☕!
One hypothesis is that the explicit coarse-graining hurts expressivity, which is preserved when modeling long-range correlations implicitly via the vanilla MPGNN. But not sure -- maybe it's possible to set-up a PointNet-style model to do better!
Yep, there are long range correlations in the data, and the global parameter $\sigma_8$ should be especially sensitive to them. Indeed PointNet++ doesn't seem to do very well at capturing them despite explicitly having hierarchical downsampling.
are we getting the greatest hits
Fun paper led by Julia Balla: "A Cosmic-Scale Benchmark for Symmetry-Preserving Data Processing".
Julia will be presenting the paper at LoG on Thursday as a spotlight oral, and also at the NeurReps Workshop at NeurIPS Workshop next month.
📄: arxiv.org/abs/2410.20516
💻: github.com/smsharma/eqn...
Would like to join!
blue sky
