Fabian Sinz

🧠Introducing OmniMouse: One of the largest datasets in neuroscience along with a systematic study of scaling properties of brain models Co-led by 🤩 K. Willeke, @pollytur.bsky.social and @alexrgil14.bsky.social Models trained on up to 3M neurons, >150B tokens

Special thanks to Paul Fahey as the lead of the entire data collection Tom Olschewski, @pollytur.bsky.social and Konstantin Willeke, for leading the effort of releasing the the data.

🧠Introducing OmniMouse: One of the largest datasets in neuroscience along with a systematic study of scaling properties of brain models Co-led by 🤩 K. Willeke, @pollytur.bsky.social and @alexrgil14.bsky.social Models trained on up to 3M neurons, >150B tokens

🎮 Interactive Visualizations Our release blog has plenty of interactive demos: toggle stimulus & behavior inputs, vary population context, and watch OmniMouse's predictions against real neural responses. enigma-brain.github.io/omnimouse

🧠Introducing OmniMouse: One of the largest datasets in neuroscience along with a systematic study of scaling properties of brain models Co-led by 🤩 K. Willeke, @pollytur.bsky.social and @alexrgil14.bsky.social Models trained on up to 3M neurons, >150B tokens

Data scaling tells the opposite story. Nested subsets of 8 → 16 → 32 → 64 → 323 sessions: performance improves monotonically for every task and every model size.

🧠Introducing OmniMouse: One of the largest datasets in neuroscience along with a systematic study of scaling properties of brain models Co-led by 🤩 K. Willeke, @pollytur.bsky.social and @alexrgil14.bsky.social Models trained on up to 3M neurons, >150B tokens

For neural response prediction (forecasting, population prediction, stimulus-conditioned encoding), gains level off around ~80M parameters. Loss curves saturate or start to overfit. At this data scale, model capacity doesn't appear to be the limiting factor.

🧠Introducing OmniMouse: One of the largest datasets in neuroscience along with a systematic study of scaling properties of brain models Co-led by 🤩 K. Willeke, @pollytur.bsky.social and @alexrgil14.bsky.social Models trained on up to 3M neurons, >150B tokens

OmniMouse is a multi-modal, multi-task transformer: single neuron tokenization, a ViT for videos, and behavior traces are fed into a unified model. Through flexible masking, our model supports neuronal encoding & decoding, forecasting, or any combination of the three.

🧠Introducing OmniMouse: One of the largest datasets in neuroscience along with a systematic study of scaling properties of brain models Co-led by 🤩 K. Willeke, @pollytur.bsky.social and @alexrgil14.bsky.social Models trained on up to 3M neurons, >150B tokens

The dataset that we’re releasing today is among the largest ever released in neuroscience: • 3.1M segmented neurons • 78 mice • 328 sessions • 150B+ neural tokens • Naturalistic + parametric visual stimuli • Behavioral data

🧠Introducing OmniMouse: One of the largest datasets in neuroscience along with a systematic study of scaling properties of brain models Co-led by 🤩 K. Willeke, @pollytur.bsky.social and @alexrgil14.bsky.social Models trained on up to 3M neurons, >150B tokens

Collaboration between the labs of @andreastolias.bsky.social , @aecker.bsky.social and @sinzlab.bsky.social at Stanford and @uni-goettingen.de Huge thanks to the entire incredible team 🤩

My lab at U Göttingen is looking for a PostDoc or PhD student to work on an ERC-funded project at the intersection of computer vision, graphics & neuroscience.

Interested? Please send your CV and a meaningful letter of motivation to sinz@uni-goettingen.de We welcome applications for women and other underrepresented groups in MINT/STEM fields.

My lab at U Göttingen is looking for a PostDoc or PhD student to work on an ERC-funded project at the intersection of computer vision, graphics & neuroscience.

The role: build computer vision methods and workflows to create realistic virtual replicas of experimental environments that allow us to render visual input from tracked animal behavior, and integrate this with deep learning models of the mouse visual system.

Early career researcher: Are you thinking about writing an ERC , Emmy Noether Proposal, or Lower Saxony AI Group application, and are you interested to work with us at @uni-goettingen.de and @goettingen-campus.de

Then register for our information event on 24. Feb and learn how we can support you to write a proposal with us. More info and registration here: www.uni-goettingen.de/en/689240.html

🔬 We're hiring a Junior Group Leader for Data-Driven Digital Twins/System Models in medicine/life sciences at @uni_goettingen! Perfect opportunity for early-career researchers who want to: - Lead their own research group - Work at the intersection of AI, simulation, and biomedicine

We offer - own funded position - 2 funded PhD positions - research budget - great collaborations (Max Planck, medical center, CIDAS) The position is part of the Lower Saxony Center for AI & Causal Methods in Medicine (caimed.de).

🔬 Exciting PostDoc Opportunity! 🐁🧠 We - the @sinzlab.bsky.social (sinzlab.org) and @trose-neuro.bsky.social (troselab.de) Labs - are seeking an experimental postdoc to work at @unibonn.bsky.social with cutting-edge miniature 2-photon microscopy and gaze tracking in freely behaving mice.

Perfect for someone who loves: ✨ Advanced imaging techniques ✨ Interdisciplinary science ✨ Understanding brain computation Interested? Please send your application (CV, motivation letter) to ​​sinz@uni-goettingen.de #neurosky #neuroskyence #Neuroscience #PostDoc #JobOpening

🔬 Exciting PostDoc Opportunity! 🐁🧠 We - the @sinzlab.bsky.social (sinzlab.org) and @trose-neuro.bsky.social (troselab.de) Labs - are seeking an experimental postdoc to work at @unibonn.bsky.social with cutting-edge miniature 2-photon microscopy and gaze tracking in freely behaving mice.

Really exciting about this work that just got accepted to ICLR as an oral. We use implicit networks to find stimulus invariance manifolds and then group neurons by aligning them to each other. Very fruitful collaboration lead by @neuralbash.bsky.social and @lucabaroni.bsky.social.