Lightnews — Scholar-powered news

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · May 5

On Incorporating Scale into Graph Networks

Standard graph neural networks assign vastly different latent embeddings to graphs describing the same physical system at different resolution scales. This precludes consistency in applications and...

openreview.net

1 2

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · May 5

The ICLR 2025 MLMP Best Paper Award, along with 2k GPU-hours from Nebius, goes to "On Incorporating Scale into Graph Networks"! Congratulations, Christian Koke, Yuesong Shen,
@abhi-rf.bsky.social, Marvin Eisenberger, @pseudomanifold.topology.rocks, Michael M. Bronstein, @dcremers.bsky.social!

Standard graph neural networks assign vastly different latent embeddings to graphs describing the same physical system at different resolution scales. This precludes consistency in applications and prevents generalization between scales as would fundamentally be needed in many scientific applications. We uncover the underlying obstruction, investigate its origin and show how to overcome it.

1 2 6

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · May 5

Paper: openreview.net/forum?id=Eb7...

ViNE-GATr: scaling geometric algebra transformers with virtual...

Equivariant neural networks can effectively model physical systems by naturally handling the underlying geometric quantities and preserving their symmetries, but scaling them to large geometric...

openreview.net

1

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · May 5

ICLR 2025 MLMP best poster award goes to "ViNE-GATr: scaling geometric algebra transformers with virtual nodes embeddings"! Congratulations @sukjulian.bsky.social, Thomas Hehn, @arashbehboodi.bsky.social, Gabriele Cesa!

Equivariant neural networks can effectively model physical systems by naturally handling the underlying geometric quantities and preserving their symmetries, but scaling them to large geometric data remains challenging. Naive downsampling typically disrupts features’ transformation laws, limiting their applicability in large scale settings. In this work, we propose a scalable equivariant transformer that efficiently processes geometric data in a coarse-grained latent space while preserving E(3) symmetries of the problem. In particular, by building on the Geometric Algebra Transformer (GATr) and PerceiverIO architectures, our method learns equivariant latent tokens which allow us to decouple the processing complexity from the input data representation while maintaining global equivariance.

1 1 4

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Apr 3

Moved to OpenReview: openreview.net/group?id=ICL...

ICLR 2025 Workshop MLMP-IRT

Welcome to the OpenReview homepage for ICLR 2025 Workshop MLMP-IRT

openreview.net

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Apr 1

📢 Machine Learning Multiscale Processes ICLR 2025 Deadline Extended to Apr 14! 📅

The good thing it's the Irreproducible Results Track, so the less experiments you run, the greater is the acceptance chance!

Submit here: forms.gle/itcBzomUsLwf...

#ScienceFail #IrreproducibleResults #AcademicHumor

forms.gle

1

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Mar 18

Authors: Marimuthu Kalimuthu, @dholzmueller.bsky.social, @mniepert.bsky.social
Full text: openreview.net/forum?id=OCM...

LOGLO-FNO: Efficient Learning of Local and Global Features in...

Learning local features and high frequencies is an important problem in Scientific Machine Learning. For instance, effectively modeling turbulence (e.g., $Re=3500$ and above) depends on accurately...

openreview.net

1 3

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Mar 18

🚀Continuing the spotlight series with the next @iclr-conf.bsky.social MLMP 2025 Oral presentation!
📝LOGLO-FNO: Efficient Learning of Local and Global Features in Fourier Neural Operators
📷 Join us on April 27 at #ICLR2025!
#AI #ML #ICLR #AI4Science

1 2 2

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Mar 12

Authors: Gianluca Galletti, Fabian Paischer, Paul Setinek, William Hornsby, Lorenzo Zanisi, Naomi Carey, Stanislas Pamela, @jobrandstetter.bsky.social
Full text: openreview.net/forum?id=SGg...

5D Neural Surrogates for Nonlinear Gyrokinetic Simulations of...

Nuclear fusion plays a pivotal role in the quest for reliable and sustainable energy production. A major roadblock to achieving commercially viable fusion power is understanding plasma turbulence...

openreview.net

1

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Mar 12

🚀Continuing the spotlight series with the next @iclr-conf.bsky.social MLMP 2025 Oral presentation!
📝5D Neural Surrogates for Nonlinear Gyrokinetic Simulations of Plasma Turbulence
📷 Join us on April 27 at #ICLR2025!
#AI #ML #ICLR

1

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Mar 10

Authors: Fang Sun, Zijie Huang, Yadi Cao, Xiao Luo, Wei Wang, Yizhou Sun (I wish at least some had a Blusky account, but alas)
Full text: openreview.net/forum?id=T86...

DoMiNO: Down-scaling Molecular Dynamics with Neural Graph Ordinary...

Molecular dynamics (MD) simulations are crucial for understanding and predicting the behavior of molecular systems in biology and chemistry. However, their wide adoption is hindered by two main...

openreview.net

1 1

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Mar 10

🚀 Next @iclr-conf.bsky.social 2025 Machine Learning Multiscale Processes Workshop oral presentation🥳
📝DoMiNO: Down-scaling Molecular Dynamics with Neural Graph Ordinary Differential Equations
📅 Join us on April 27 at #ICLR2025!
📷 Early reg. deadline: March 15 #AI #ML #ICLR

1 1 2

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Mar 8

Paper full text: openreview.net/forum?id=SRC...

On the successful Incorporation of Scale into Graph Neural Networks

Standard graph neural networks assign vastly different latent embeddings to graphs describing the same physical system at different resolution scales. This precludes consistency in applications and...

openreview.net

1 2

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Mar 8

Authors: Christian Koke, Yuesong Shen, @abhi-rf.bsky.social, Marvin Eisenberger, @pseudomanifold.topology.rocks, Michael M. Bronstein, @dcremers.bsky.social

1 1 2

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Mar 8

🚀 Excited to kick off our @iclr-conf.bsky.social 2025 Machine Learning Multiscale Processes Workshop contributed paper series! 🥳

📝 On the Successful Incorporation of Scale into Graph Neural Networks

📅 Join us on April 27 at #ICLR2025!
⏳ Early reg. deadline: March 15

#AI #ML #ICLR

1 2 3

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Mar 6

🚨 Paper decisions for the @iclr-conf.bsky.social 2025 Workshop on Machine Learning for Multiscale Processes are out!

👉 Check out the latest in multiscale ML openreview.net/group?id=ICL...

Save the date: April 27, 2025! 🗓️

Stay tuned—we’ll be spotlighting accepted papers soon! 🌟

ICLR 2025 Workshop MLMP

Welcome to the OpenReview homepage for ICLR 2025 Workshop MLMP

openreview.net

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Feb 13

📝 Start writing now, submit on Feb 23! If we solve scale transition, we solve science.

#ICLR2025 #MachineLearning #MultiscaleModeling #AIforScience

1

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Feb 13

💻 GPU grants from our industry partners Constructor Tech and Nebius.

1

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Feb 13

🌏 “Last call” to attend @iclr-conf.bsky.social and visit Singapore. Short paper track is specifically designed to be accessible to budding researchers; ICLR offers financial assistance. Singapore is the most expensive city in the world – because it's worth it!

1

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Feb 13

🔬Top science. The event is part of the №2 AI conference in the world; keynotes by Singapore's only Nobel Laureate, Kostya Novoselov ∈ National University of Singapore, Sergei Gukov ∈ @caltech.edu, Charlotte Bunne ∈ @icepfl.bsky.social, Qianxiao Li ∈ NUS, Daniel Polani ∈ U of Hertfordshire.

1

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Feb 13

📢 Machine Learning Multiscale Processes @iclr-conf.bsky.social 2025 Deadline Extended to Feb 23! 📅

Given low-level theory and computationally-expensive simulation code, how can we model complex systems on a useful time scale?

Why attend? 🧵

1 2

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Feb 3

Thank you for reading though! We (the multiscale workshop) now have a website, and would like to invite you to update the board) multiscale-ai.github.io

Machine Learning Multiscale Processes @ ICLR 2025

multiscale-ai.github.io

1 1

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Jan 28

🚀💻 Interested in Artificial General Intelligence?

Want to hear more about how Reinforcement Learning could solve Millennium Prize problems?

Attend our workshop and listen to @caltech.edu Prof. Sergei Gukov's talk: Math + AI = AGI 🌟

Submit your work here multiscale-ai.github.io!

Machine Learning Multiscale Processes @ ICLR 2025

multiscale-ai.github.io

Reposted by Multiscale AI @ ICLR 2025

ICLR Nucleic Acids Workshop @ai4na-workshop.bsky.social · Jan 27

🔥If you think that AI4NA is not a perfect fit for your project, make sure to check out other awesome AI4Science workshops: AgenticAI, @gembioworkshop.bsky.social, @multiscaleai.bsky.social, @climatechangeai.bsky.social, MLGenX

1 2 2

Multiscale AI @ ICLR 2025 @multiscaleai.bsky.social · Jan 23

💥Call for Submissions for the ICLR2025 @iclr-conf.bsky.social Workshop on ML for Multiscale Processes is open💥

Submission deadline is on the 10th of February!
Please submit here: multiscale-ai.github.io

Looking forward to receiving your papers!!!

Machine Learning Multiscale Processes @ ICLR 2025

multiscale-ai.github.io

1