🪴Node-Level Topological Representation Learning on Point Clouds🪴
My new preprint with Michael Schaub relating global topology with node-level features is now on arxiv! 🎉
https://t.co/fwOGEKzIJr
A Python package is available as well. (I think it's pretty cool. 😊) 👇👇
We have turned the paper into a python package! Check out TOPF🪴with example jupyter notebooks at
https://t.co/Q2wVAGSDVM
or simply pip install it via
pip install topf
Our paper “Point-Level Representation Learning on Point Clouds” with Michael Schaub has been accepted at ICML 2025! 🎉
We link global topological descriptors back to individual points — connecting TDA & local geometry.
https://t.co/fwOGEKzaTT
See you in Vancouver next week!🇨🇦
I can understand that all reviewers favouring accept over reject does not lead to an accept and can live with this decision. But there needs to be a way to improve the system in the future that at least the decisions seem to be ... based on facts and the actual reviews? 2/n
After a great keynote this morning by @KrishnaswamyLab, I had a lot of fun presenting my work on Hodge Learning on Point Clouds at the Graph Signal Processing Workshop at @TUDelft_AI . Thank you very much to the organisers and the audience! 😊
More great talks at GSP on higher order interaction learning. Also a talk by our own Charles Xu on hyperedge representations. Also managed to squeeze in a visit to Royal Delft, the historic blue ceramic made here!
In "Hodge Learning: What the Eigenspectrum of the Hodge Laplacian tells us about higher-order network topology and geometry", I will talk about applying ideas from differential geometry to point clouds and higher-order networks. Come by! 😊
I had a wonderful first day at @NetSciConf in Québec with great satellites on topological network science @TopoNets and statistical inference. 🎉
Today, I am excited to present my work on Hodge Learning at the Higher-Order Network Models Satellite at 5:30 pm in room 206A. 😊
@MEGALOCER0S Well I don't have one, so ... hopefully not? 😅 You need topologically structured data though (Or the examples of the Jupyter notebook in github 😂
🪴Topological Point Features (TOPF)🪴
If you're running Linux or macOS, simply install Julia and use `pip install topf` for topological point features and extensive visualisation option.
For two example notebooks, check out the github page!
https://t.co/Q2wVAGTbLk
Excited to be giving a talk on node-level topological representations on point cloud and my newest paper (🪴) at the @helmholtz_ai conference in Düsseldorf this afternoon at 16:00! If you are interested in what algebraic topology can do for ML and AI, see you in room 3! 🎉 #TDA
@rballeba Good luck! And same for me this year... But so far, 5/6 papers I am assigned sound really interesting and I would have wanted to read them anyway. 🎉 I really did not expect this!
🪴Node-Level Topological Representation Learning on Point Clouds🪴
My new preprint with Michael Schaub relating global topology with node-level features is now on arxiv! 🎉
https://t.co/fwOGEKzIJr
A Python package is available as well. (I think it's pretty cool. 😊) 👇👇
Half an hour left of my poster presentation of my #ICASSP2024 paper „Disentangling the Spectral Properties of the Hodge Laplacian“ with Michael Schaub! Talk to me at Poster zone 6C! 🎉
@ieeeICASSP
I‘ll be at ICASSP 2024 in Seoul until the end of week. If anyone is interested in talking about Signal Processing over (higher-order) Networks, Topological Signal Processing (or has tips on where to find vegetarian food in Korea :D), feel free to reach out! :)
@ieeeICASSP