Lightnews — Scholar-powered news

Reposted by Erik Bekkers

Julian Suk @sukjulian.bsky.social · May 24

Enschede crowd you don't want to miss this! Join us for two exciting talks by Jolanda Wentzel and @erikjbekkers.bsky.social. As a bonus, you get to see me in "appropriate traditional dress from [my] country of origin".

📅 28 May
📍 UTwente campus

5 11

Reposted by Erik Bekkers

Sara Magliacane @smaglia.bsky.social · May 12

New PhD position at the University of Amsterdam in @amlab.bsky.social on learning concepts with theoretical guarantees using #causality and #RL with me, Frans Oliehoek (TU Delft) and Herke van Hoof 💥

Deadline: 15 June

werkenbij.uva.nl/en/vacancies...

Vacancy — PhD Position on Learning Concepts with Theoretical Guarantees Using Causality and RL

Are you interested in improving the interpretability, robustness and safety of current AI systems? If the answer is yes, please continue reading!

werkenbij.uva.nl

9 22

Reposted by Erik Bekkers

ELLIS @ellis.eu · Mar 10

💬 Join TODAY’s webinar: “Symmetry, scale, and science: A geometric path to better AI" by @erikjbekkers.bsky.social & @jobrandstetter.bsky.social
(ELLIS Program “Geometric Deep Learning”)

🗓️ Mon, March 10, 2025
🕓 16:00-17:00 CET

🔗 Register here: aiforgood.itu.int/event/symmet...

Symmetry, scale, and science: A geometric path to better AI

The success of modern AI systems has been largely driven by massive scaling of data and compute resources. However, in scientific applications, where

aiforgood.itu.int

2 9

Reposted by Erik Bekkers

Olga Zaghen @olgatticus.bsky.social · Mar 6

Super happy to share our work was accepted as an Oral at the Delta Workshop @ ICLR 2025! 🎉

Can’t wait to talk about it in Singapore 😎

Congrats to the amazing team @eijkelboomfloor.bsky.social @alisometry.bsky.social @erikjbekkers.bsky.social 🔥

Olga Zaghen @olgatticus.bsky.social · Feb 19

Variational Flow Matching goes Riemannian! 🔮

In this preliminary work, we derive a variational objective for probability flows 🌀 on manifolds with closed-form geodesics, and discuss some interesting results.

Dream team: Floor, Alison & Erik (their @ below) 💥

📜 arxiv.org/abs/2502.12981
🧵1/5

4 13

Erik Bekkers @erikjbekkers.bsky.social · Feb 26

🚀 Excited to be part of this project led by Praneeta Konduri! We’re using state-of-the-art generative models to create synthetic vasculature geometries—pushing stroke treatment development forward while cutting down on patient data reliance. Exciting stuff! 😃 rdt.uva.nl/research/res...

Synthetic data to enhance new treatment uptake for acute ischemic stroke - Responsible Digital Transformations

Multiple (pre-)clinical trials are required in the regulatory trajectory to introduce new medical devices in clinical practice. The process is widely acknowledged as sub-optimal, resource-heavy, and t...

rdt.uva.nl

1 7

Reposted by Erik Bekkers

Olga Zaghen @olgatticus.bsky.social · Feb 19

Variational Flow Matching goes Riemannian! 🔮

In this preliminary work, we derive a variational objective for probability flows 🌀 on manifolds with closed-form geodesics, and discuss some interesting results.

Dream team: Floor, Alison & Erik (their @ below) 💥

📜 arxiv.org/abs/2502.12981
🧵1/5

1 12 34

Reposted by Erik Bekkers

Georg Bökman @bokmangeorg.bsky.social · Feb 10

Common beliefs about equivariant networks for image input include 1) They are slow. 2) They don’t scale to ImageNet. 3) They are complicated. In my opinion, these three are all false. To argue against them, we made minimal modifications to popular vision models, turning them mirror-equivariant.

2 4 26

Reposted by Erik Bekkers

Christian A. Naesseth @canaesseth.bsky.social · Feb 5

Really excited about this! We note a connection between diffusion/flow models and neural/latent SDEs. We show how to use this for simulation-free learning of fully flexible SDEs. We refer to this as SDE Matching and show speed improvements of several orders of magnitude.

arxiv.org/abs/2502.02472

SDE Matching: Scalable and Simulation-Free Training of Latent Stochastic Differential Equations

The Latent Stochastic Differential Equation (SDE) is a powerful tool for time series and sequence modeling. However, training Latent SDEs typically relies on adjoint sensitivity methods, which depend ...

arxiv.org

10 50

Reposted by Erik Bekkers

Cees Snoek @cgmsnoek.bsky.social · Feb 3

✨ The VIS Lab at the #University of #Amsterdam is proud and excited to announce it has #TWELVE papers 🚀 accepted for the leading #AI-#makers conference on representation learning ( #ICLR2025 ) in Singapore 🇸🇬. 1/n
👇👇👇 @ellisamsterdam.bsky.social

1 4 17

Reposted by Erik Bekkers

Artem Moskalev @artemmoskalev.bsky.social · Feb 3

Accepted to ICLR 🚨 Does using more geometry always help with molecule property prediction? In practice, we deal with imperfect geometries, which introduce structural noise.

In our work arxiv.org/abs/2410.11933, we investigate when and how geometric information is useful (or not) for RNA molecules.

Beyond Sequence: Impact of Geometric Context for RNA Property Prediction

Accurate prediction of RNA properties, such as stability and interactions, is crucial for advancing our understanding of biological processes and developing RNA-based therapeutics. RNA structures can ...

arxiv.org

1 1 13

Reposted by Erik Bekkers

Efstratios Gavves @egavves.bsky.social · Jan 23

🔥 3/4 #ICLR2025 Grounding Continuous Representations in Geometry: Equivariant Neural Fields (ENF), for geometry-informed continuous signal representations openreview.net/forum?id=A4e...
[email protected], @davidmknigge.bsky.social @erikjbekkers.bsky.social S. Papa, R. Valperga, S. Vadgama

Grounding Continuous Representations in Geometry: Equivariant...

Conditional Neural Fields (CNFs) are increasingly being leveraged as continuous signal representations, by associating each data-sample with a latent variable that conditions a shared backbone...

openreview.net

1 8

Reposted by Erik Bekkers

ELLIS @ellis.eu · Jan 29

🚀 AI for Good Webinar Launch: From Molecules to Models

Next week, ELLIS kicks off a new webinar series on AI in life sciences, showcasing key research from ELLIS Programs.

🗓️ Date: Feb 3, 2025
🕓 Time: 16:00-17:30 CET

Register: aiforgood.itu.int/event/unlear...

Unlearning Toxicity in Multimodal Foundation Models & Learning to design protein-protein interactions with enhanced generalization

Part 1 (Rita Cucchiara): Foundation Models, pretrained on extremely large unknown source of data, contain in their embed space many information

aiforgood.itu.int

3 6

Reposted by Erik Bekkers

AMLab @amlab.bsky.social · Dec 9

♻️Learning Symmetries via Weight-sharing with Doubly Stochastic Tensors

by Putri van der Linden, @algarciacast.bsky.social, @sharvaree.bsky.social, Thijs P. Kuipers, @erikjbekkers.bsky.social

🪪https://neurips.cc/virtual/2024/poster/96699
📜https://arxiv.org/abs/2412.04594

🧵9/ 12

Learning Symmetries via Weight-Sharing with Doubly Stochastic Tensors

Group equivariance has emerged as a valuable inductive bias in deep learning, enhancing generalization, data efficiency, and robustness. Classically, group equivariant methods require the groups of in...

arxiv.org

1 2 10

Reposted by Erik Bekkers

Floor Eijkelboom @eijkelboomfloor.bsky.social · Dec 6

🐑 Come and check out Variational Flow Matching for Graph Generation next week at @neuripsconf.bsky.social ! 🐑

Wed 11 Dec 11 a.m. PST — 2 p.m. PST
West Ballroom A-D #7103

arxiv.org/abs/2406.04843

4 26

Reposted by Erik Bekkers

AMLab @amlab.bsky.social · Nov 21

Meet our Lab's members: staff, postdocs and PhD students! :)

With this starter pack you can easily connect with us and keep up to date with all the member's research and news 🦋

go.bsky.app/8EGigUy

1 9 25

Reposted by Erik Bekkers

Olga Zaghen @olgatticus.bsky.social · Nov 20

A starter pack for researchers interested in Geometric Deep Learning - in the broadest sense possible!

Let me know if you would like to be listed. :)

Thanks @sharvaree.bsky.social for the idea!

go.bsky.app/7h8sek

34 20 98

Reposted by Erik Bekkers

AMLab @amlab.bsky.social · Nov 19

Soon, @erikjbekkers.bsky.social and @davidmknigge.bsky.social will give a talk elaborating even further on geometry-grounded representation learning in a NeurReps seminar. Make sure to mark the date! :)

⏰ November 21st, 4 PM CET
🔗 www.neurreps.org/speaker-seri...

1 5 18