Steffen Schneider (@stes_io) header
Steffen Schneider (@stes_io) profile picture

Research Group Leader @HelmholtzMunich. 🦓@CebraAI. Co-Founder @KImachtSchule & @KinematikAI. ➡️ Hiring PhD students & SWEs!

Munich, Germany 🥨
Joined October 2017
Photo by Steffen Schneider (@stes_io)Photo by Steffen Schneider (@stes_io)Photo by Steffen Schneider (@stes_io)Photo by Steffen Schneider (@stes_io)Photo by Steffen Schneider (@stes_io)Photo by Steffen Schneider (@stes_io)Photo by Steffen Schneider (@stes_io)Photo by Steffen Schneider (@stes_io)Photo by Steffen Schneider (@stes_io)Photo by Steffen Schneider (@stes_io)
Beyond excited that @cebraAI is out in @Nature today: A new algorithm leveraging self-supervised and supervised contrastive learning for scientific discovery & hypothesis testing! 📑doi.org/10.1038/s41586-023-0… 🦓 cebra.ai Read more below🔽🔽🔽
🦓 Self-supervised multimodal ML is promising the next AI breakthrough - in our new work published in @Nature, we debut @CEBRAai: for self-supervised hypothesis- and discovery-driven science. 📝 doi.org/10.1038/s41586-023-0… 💻github.com/AdaptiveMotorCont… 🦓 cebra.ai/ 🧵⬇️
5
14
1
93
16,073
Steffen Schneider retweeted
🚀 We released a new version! v0.3.1 includes a new @plotlygraphs integration, which is very useful for embedding exploration in notebooks 💜🔥🚀🦓
1
12
69
6,907
Steffen Schneider retweeted
We proudly present our 524 page book on equivariant convolutional networks. Coauthored by Patrick Forré, @erikverlinde and @wellingmax. maurice-weiler.gitlab.io/#cn… [1/N]
27
241
23
1,120
161,601
Steffen Schneider retweeted
📢Our work on compositional generalization has been accepted to NeurIPS 2023! Compositional generalization is still very hard for machines and doesn't emerge in even simple regression settings. We analyze this theoretically and show when it can occur. 1/7 arxiv.org/abs/2307.05596
2
37
2
201
36,936
Steffen Schneider retweeted
Exciting! There is a new demo notebook for how to use #CEBRA on @AllenInstitute OpenScope project data! 🎉 contributed by @LecoqJerome, now live at cebra.ai/docs/demos.html #opensourcedata 🦓🫶 #machinelearning4neuro
1
8
1
67
13,505
It was great to discuss @cebraAI at #NeuroDataReHack2023. Thanks for hosting me, @NeurodataWB !
Steffen @stes_io from @TrackingActions group walks #NeuroDataReHack2023 participants through the #CEBRA method for learning latent embeddings for joint behavioral and neural analysis and how to apply it to NWB data from macaque M1 during reaching dandiarchive.org/dandiset/00…
12
851
🔔Updated on arXiv: it’s a pretty big overhaul of presentation and new SOTA results, plus even better model performance for quadrupeds! #PoseEstimation #ComputerVision #neuroscience #MachineLearning arxiv.org/abs/2203.07436
🐁🐘🐿🔥SuperAnimals for #Pose -No human labeling - Video analysis on over 45 species with only 2 global classes of animal pose models -If SA needs fine-tuning, its 10× more data efficient & 2X outperforms prior transfer learning 📝@shaokaiyeah et al arxiv.org/abs/2203.07436 🧵⬇️
23
1
106
33,503
Steffen Schneider retweeted
This week at #ICML I will present “Hyperbolic Image-Text Representations” — we bring hyperbolic geometry to CLIP and learn interpretable models that also match/outperform CLIP on standard vision tasks. 📜 arxiv.org/abs/2304.09172 📅 Jul 27, 10:30. Say hi! With @MetaAI @michigan_AI
20
81
16,213
🧠🐭🔭 Neuroscience is in a new era: we’ve never been able to collect as much neural & behavior data as we can now,🔋 by new eng & AI. We aim to build a Brain & Behavior Observatory, building on community #WeNeedYou 📝: pubmed.ncbi.nlm.nih.gov/3624… 🧠🐭🔭: BrainBehaviorObservatory.org
3
90
4
290
50,087
🔮Ever wished you could just say what behavioral analysis you wanted? ... "Plot the animal's trajectory" "Count the head hips in the open arm" 👀🪄we present AmadeusGPT🎻 to do this ... 🗞@shaokaiyeah @jessy_lauer @zhoumu53 @trackingplumes and me  arxiv.org/abs/2307.04858 🧵⬇️
7
186
37
608
196,827
Steffen Schneider retweeted
We are excited to announce that Europe's first ELLIS Institute is launched in Tübingen, Germany! It offers outstanding research conditions for top researchers from all over the world who will work together on challenges & opportunities of #AI. ellis.eu/news/europe-s-first…
1
44
1
201
28,049
🚨Crowded scenes are notoriously difficult for #ComputerVision models. In our new work lead by @zhoumu53 & @LStoffl we developed a novel approach called BUCTD that is state-of-the-art on crowded 🐒🐠&🕺human pose estimation benchmarks!🔥 arxiv.org/abs/2306.07879 #CVPR2023 workshop
6
67
7
342
67,245
It was great to speak about @cebraAI at the @helmholtz_ai conference in Hamburg!
Our Session 3a brought together excellent speakers on the topic "Probably explainable". It was a privilege to hear @stes_io, Martin Dyrba, Stefano Bianco, Benjamin Schäfer, Mario Teixeira, and @vincefort! #HelmholtzAICon #ML #AI @desynews
1
16
1,597
Steffen Schneider retweeted
In our latest work (oral @ #ICML2023 🎉), we study when learning object-centric representations is theoretically possible and identify two surprisingly intuitive assumptions on the data-generating process: compositionality and irreducibility (see 🧵). 1/8 brendel-group.github.io/obje…
1
55
3
226
48,409
Check out our latest release of 🦓cebra on PyPI! Easy-to-use contrastive learning for analysing datasets in neuroscience & beyond! 🦾🧠📊 🔽 Here's how to get started 🔽
#cebra is out on pypi & available for research use! Check out the amazing docs, demos, & plz ⭐️ the repo!🎉 🦓pip install cebra ⭐️code: github.com/AdaptiveMotorCont… 📚docs/demos: cebra.ai/docs 📝nature.com/articles/s41586-0… 👥@stes_io @jinnnnnlee @BenquetCelia @TrackingActions
1
5
32
4,325
Steffen Schneider retweeted
Brain dynamics uncovered using a machine-learning algorithm go.nature.com/3NCMdz2
1
25
111
16,750
🦓 Self-supervised multimodal ML is promising the next AI breakthrough - in our new work published in @Nature, we debut @CEBRAai: for self-supervised hypothesis- and discovery-driven science. 📝 doi.org/10.1038/s41586-023-0… 💻github.com/AdaptiveMotorCont… 🦓 cebra.ai/ 🧵⬇️
30
266
39
976
251,510
Steffen Schneider retweeted
New paper from @TrackingActions! I liked this one because neuroscience has seen huge advances in recording technologies and behavioral tracking, but it's not always clear what to do with all the data. This paper shows an elegant approach nature.com/articles/s41586-0…
4
68
4
263
55,252
Steffen Schneider retweeted
Really excited to present new work by @ritagonmar: we visualized the entire PubMed library, 21 million biomedical and life science papers, and learned a lot about -- THE LANDSCAPE OF BIOMEDICAL RESEARCH biorxiv.org/content/10.1101/… Joint work with @CellTypist and @benmschmidt. 1/n
28
372
63
1,270
797,014