Tomorrow (Monday) reading group session: Michael Galkin presents "Towards Foundation Models for Knowledge Graph Reasoning" arxiv.org/abs/2310.04562! I predict there will be at least 1 meme slide🙃 Join us on Zoom at 11am ET / 5pm CET / 4pm UTC: portal.valencelabs.com/logg

This arxiv.org/abs/2402.05501 looks like a nice overview of the current landscape of using machine learning to enhance mixed-integer linear optimization problems. (By Lara Scavuzzo, Karen Aardal, Andrea Lodi, Neil Yorke-Smith.)

6 papers to review for ICML Conference in less than 30 days. This is not sustainable, especially because these are all either deeply theoretical work on geometry, or completely new applied work (not we-beat-sota-by-1%). I’ll still do my best job but my own work will suffer.

A young student wants to start a side-project with music machine learning (would be their first ML project), What's a good starting point in terms of datasets that would work with modern methods?

Question for LLM friends: Do we know what self-supervised objective was used to train Gemma (or Llama, Mistral for that matter)? Is it just autoregressive next token prediction at scale, or are more advanced span denoising objectives (eg. UL2) what top labs use right now?

Come and join us: great people, exciting problems, large datasets and now a powerful cluster 🔥

Come to hear about our work! Today (Mon) 6pm Israel time on zoom: portal.valencelabs.com/logg

The headline for AlphaFold3 is no equivariance. But there’s so much more to the methods that is challenging and thought provoking.

Thanks Michael Bronstein — such a nice adventure it has been! P.S. Guadalupe Gonzalez , attaching a little memory from the graduation day…

Debate on equivariance: I feel that ML potentials cannot be framed within the same AF3 discussions about equivariance. In fact, these discussions have triggered quite fundamental questions on my part. I am really eager to hear from other people 🧵:

We are hosting a Graph Learning Social ICML Conference on Thursday 5:30pm in room Stolz 1! 🔥 Highlights: * Snacks and drinks provided (thanks to AstraZeneca) 🍔 * Panel on the future and challenges of graph learning hosted by Bastian Grossenbacher-Rieck 💬 * Feel free to bring your poster 🖼️

Next we have «The Expressive Power of Path-Based Graph Neural Networks» by Caterina Graziani Tamara Drucks Fabian Jogl #ICML2024

Come to our Graph Learning Social today ICML Conference at 5:30pm in room Stolz 1! Tentative panelists: Christopher Morris Haggai Maron Fabrizio Frasca Derek Lim Corinna Coupette and Dimitrios Athanasakis 🔥

🧵 Our geometric GRaM Workshop at ICML 2024 workshop has released the accepted blog posts and tutorials! Theory can be hard to digest, and educational materials are scarce. We are impressed by the students' work! gram-blogposts.github.io Some of my favorites ⬇️

Beautiful ❤️ I'm really looking forward to it. The paper is such a nearly organized perspective on the field. Helpful to newcomers as well as the seasoned GDLers to get inspiration and put their work in perspective!! We're in for a treat GRaM Workshop at ICML 2024 ! 🤗

Friends, I am beyond happy! I'm starting a new position as Full Professor of #MachineLearning at the University of Fribourg Uni Fribourg 🇨🇭! With #SwissAI and many other initiatives, I am taking my research at the intersection of #geometry, #topology, and #MachineLearning to a new

Can Topological Deep Learning (TDL) models compute fundamental topological invariants? 🔎🌐 Our new paper suggests that in many cases the answer is surprisingly NO. To mitigate this, we develop two new TDL architectures inspired by expressive graph networks. 1/11

Really excited to be part of this effort — expressivity and topology together again offering new insights 🙂

We are glad to announce that the Learning on Graphs Conference 2024 Italian meet-up will be hosted in Siena! 🇮🇹✨️ 🗓️ Join us from December 4th to 6th. 🔜 Registration will open soon! 👉 For more information visit our website : sites.google.com/u/0/d/1XCyjW-u… #log #graphs #graphlearning #ml #siena #gnn