ML conference proceedings are modern-day bestiaries

Thrilled to announce that we just presented „MAGNet: Motif-Agnostic Generation of Molecules from Scaffolds“ at #ICLR2025 🧲 Johanna Sommer Bastian Grossenbacher-Rieck Fabian Theis Stephan Günnemann For those who couldn’t make it to our spotlight: openreview.net/forum?id=5FXKg…

📣 Our spicy ICML 2025 position paper: “Graph Learning Will Lose Relevance Due To Poor Benchmarks”. Graph learning is less trendy in the ML world than it was in 2020-2022. We believe the problem is in poor benchmarks that hold the field back - and suggest ways to fix it! 🧵1/10

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, Abhishek Saroha, Marvin Eisenberger, Bastian Grossenbacher-Rieck, Michael Bronstein, Daniel Cremers!

😇Hijacking this great thread to talk about a potential way to _identify_ problems with existing benchmarking datasets (also to appear in ICML 2025): arxiv.org/abs/2502.02379 TL;DR: We look at how "graphical" graph benchmarking datasets are. Spoiler: less than you'd hope for!

It's almost time for #GSP2025 ⏱️🥳 We're super excited to host you at Mila - Institut québécois d'IA this coming Wednesday, 14th May! Hope those posters are coming along 😉 👉🏻 FYI: Registrations are still open, so we encourage you to visit our website for details: gspworkshop.org

Happy weekend and happy Mother’s Day!

the only way to get to where you want to go AND be happy is to enjoy the journey. Otherwise, you'll arrive and move the goal post again and again and again while not being happy. Perpetually never reaching this fantasy of happiness.

Why math? Because I want to. I want to realise I don't understand stuff I thought I could. That's why.

Psyched to begin our first day at Mila - Institut québécois d'IA for #GSP2025 🥳 We have an exciting program in store for you! The updated schedule is now up on our website: gspworkshop.org/schedule/ See you there 🤝🏻

embarking on a fun summer project & playing claude-3.7 and GPT-4o off each other for prose writing. much to my surprise, GPT-4o is doing a better job with tone and stylistic subtleties.

ive came so far since 2021. started not knowing how to code, straight out of high school working retail and a bar while being a carer and working my ass up from nothing. math arc is the beginning of the end, the final destination. each year i did something new: 2019-2020:

[1/4] Hello guys! We have a new preprint:⚠️Graph and Simplicial Complex Prediction Gaussian Process via the Hodgelet Representations⚠️ 🔗Link: arxiv.org/abs/2505.10877

[2/4] In this work, we develop a framework for constructing Hodgelet representations of graphs and simplicial complexes, tailored for downstream prediction tasks such as classification and regression. These representations can be easily combined with Gaussian processes to provide

[3/4] The key contribution lies in the fact that the Hodgelet representations leverage both the wavelet transform of attributes and the Hodge decomposition. The wavelet transform captures multi-scale spectral and spatial information, while the Hodge decomposition enables finer

[4/4] Please don't hesitate to reach out! This is a joint work with great researchers: So Takao xiaowen dong Bastian Grossenbacher-Rieck Emmanuel Noutahi

Keiran xiaowen dong Bastian Grossenbacher-Rieck Emmanuel Noutahi Hopefully soon!

what’s up my non-orientable manifolds

Friends from the #AcademicTwitter, one of our students Uni Fribourg is interested in studying the acceptance of LLM use in social media. Can you spare a couple of minutes to answer their survey? survey.unifr.ch/index.php/3823… A variety of responses would mean the world to us!

GLOW is coming back this Wednesday! 🌟 We will hear from – and interact with – Christian Koke (incorporating scale in GNNs) and Yonatan Sverdlov (sparse geometric MPNNs and their expressive power). 🗓️When May 28th, 5pm CEST on Zoom. 🌐 Details & sign-up: sites.google.com/view/graph-lea….