Rudy Morel
@rdMorel
Followers
261
Following
200
Media
4
Statuses
42
AI&Science | Resarch Fellow at @FlatironCCM | Member of @PolymathicAI | ex PhD student at @ENS_Ulm
New York
Joined August 2023
Diffusion models learn probability densities by estimating the score with a neural network trained to denoise. What kind of representation arises within these networks, and how does this relate to the learned density? @EeroSimoncelli @StephaneMallat and I explored this question.
12
81
480
Wanna really fundamentally change science using ML? Come join the @PolymathicAI initiative at our Cambridge (UK) location!
PolymathicAI is recruiting two postdoctoral researchers to join our team at Cambridge, to work on building and understanding large-scale foundation models for science. Please share with potential candidates!
1
4
16
Does a smaller latent space lead to worse generation in latent diffusion models? Not necessarily! We show that LDMs are extremely robust to a wide range of compression rates (10-1000x) in the context of physics emulation. We got lost in latent space. Join us 👇
14
91
465
Not sure if the physics community is aware, but Polymathic AI has been quietly putting 10TB of physics simulation data on @huggingface. ⛓️💥 https://t.co/3iB8ExlXwt
8
36
224
For evolving unknown PDEs, ML models are trained on next-state prediction. But do they actually learn the time dynamics: the "physics"? Check out our poster (W-107) at #ICML2025 this Wed, Jul 16. Our "DISCO" model learns the physics while staying SOTA on next states prediction!
4
50
302
What is the probability of an image? What do the highest and lowest probability images look like? Do natural images lie on a low-dimensional manifold? In a new preprint with @ZKadkhodaie @EeroSimoncelli, we develop a novel energy-based model in order to answer these questions: 🧵
11
74
366
What can swimming bacteria teach us about how the ocean’s layers mix? @PolymathicAI recently released two massive datasets for training artificial intelligence models to tackle problems across scientific disciplines, available on @huggingface. Learn more:
simonsfoundation.org
New Datasets Will Train AI Models to Think Like Scientists on Simons Foundation
0
6
14
Congratulations to #FlatironCCM research scientist @JiequnH on being awarded @TheSIAMNews' 2025 SIAG/CSE Early Career Prize! Read more: https://t.co/now51Ox3n6
#math
simonsfoundation.org
The Flatiron Institute’s Jiequn Han will receive the Early Career Prize from the Society of Industrial and Applied Mathematics.
0
2
18
📢Tell your friends who want to work on building large foundation models in astronomy: Application to our postdoctoral positions in Paris 🇫🇷 is closing in a few days!! The deadline is on Feb 14, 202 :)
You saw our AstroCLIP, and have heard of MultiModal Universe! Now, come build and scale Foundation models for Astrophysics with us at @PolymathicAI! Check this out: https://t.co/lcOYwX4mnX Deadline: Feb 14, 2025!
0
6
18
Many exciting questions in ML&science right now! Check out our summer internships in the Center for Computational Mathematics at the Flatiron Institute @FlatironCCM @FlatironInst Location: Manhattan, New York Apply here: https://t.co/RloQjumUOf Deadline: Jan. 15, 2025
1
7
39
Excited to be giving an invited talk at Foundation Models for Science workshop at Neurips in ~1.5 hours about @PolymathicAI ! Come join us at West Meeting Room 202-204! at 11:15am! 🔥 Thank you to the organizers for making this workshop possible! https://t.co/q83HIwlg08
5
4
70
🔥 @PolymathicAI presents the coolest (IMO) and most diverse fluid dynamics dataset The Well at #NeurIPS2024 ! Date: tomorrow (Thursday) Dec 12 Time: 11am-2pm (PST)! Where: West Ballroom A-D #5102 Led by @mikemccabe210 and @oharub !! Data + code: https://t.co/uFvtydBQe2
1
6
30
Excited to be at NeurIPS this week! 🎉 I'm part of four exciting projects being presented: The Well & Multimodal Universe: massive, curated scientific datasets LaSR: LLM concept evolution for symbolic regression MPP: 0th gen @PolymathicAI All posters Wed/Thu - stop by! 👋
0
4
34
Very proud of what we accomplished with @PolymathicAI + MultiModal Universe Collaboration! 👉100 TBs of Astronomical data ⭐️🌟💫 👉From > 10 telescopes, over 20 modalities 👉All ML training ready This will be presented today (Wed) at 4:30pm-7:30pm (PST) at West Ballroom A-D
0
9
46
🔥@PolymathicAI team will present Multiple Physics Pretraining (MPP) tomorrow from 11am to 2pm (PST) at East Exhibit Hall A-C #4100 🔥 at #NeurIPS2024! Learn how to pre-train on various incompressible fluids simulations and make the AI model predict what will happen to nearly
3
11
78
I will be attending #NeurIPS2024! Feel free to ping me or the @PolymathicAI team if you want to chat about The Well 🥤or MPP 🚤! And since you like eye-catchy scientific gifs, here is a Rayleigh-Bénard convection one 😁 🥤: https://t.co/XEUBBreOq0 🚤: https://t.co/ZTwrW0mvlO
1
3
21
Applications to our Research Fellow position at @FlatironCCM are closing soon on Dec 15! It's a great place for doing fundamental ML research with a lot of freedom in a great environment, in the heart of NYC. Apply here:
0
25
82
Neural networks trained on diverse tasks share foundational patterns, says CDS Faculty Fellow Florentin Guth (@FlorentinGuth) & @JohnsHopkins' Brice Ménard. Their findings shed light on transfer learning's success & neural networks' universal encodings. https://t.co/XJeXG9bJYk
nyudatascience.medium.com
Neural networks trained on different tasks still learn the same foundational patterns, revealing universality in their encodings.
0
4
15