Alexander Immer

In our #NeurIPS2021 paper (arxiv.org/abs/2106.14806), we introduce laplace-torch for effortless Bayesian deep learning. Despite their simplicity, we find that Laplace approximations are surprisingly competitive with more popular approaches. youtu.be/nMONiYLWWOU

1/14 Is Muon “better” than Shampoo? We argue that their relationship parallels Adam's relationship with Signum. Analogous to @lukas_balles and Hennig’s (2018) decomposition of Adam into element-wise scaled Signum, we can decompose Shampoo as left- and right-adapted Muon.

KFAC is everywhere—from optimization to influence functions. While the intuition is simple, implementation is tricky. We (@BalintMucsanyi, @2bys2 ,@runame_) wrote a ground-up intro with code to help you get it right. 📖 arxiv.org/abs/2507.05127 💻 github.com/f-dangel/kfac-…

In collaboration with HHMI Janelia & Harvard, we introduce ZAPBench, a whole-brain activity dataset and benchmark with single cell resolution of a larval zebrafish, which will enable development and comparison of more accurate brain activity models. →goo.gle/44rEpsn

How do you identify training data responsible for an image generated by your diffusion model? How could you quantify how much copyrighted works influenced the image? In our ICLR oral paper we propose how to approach such questions scalably with influence functions.

Can we predict future brain activity in a small vertebrate?🤔 We're releasing ZAPBench⚡️(#ICLR2025 spotlight): a benchmark to forecast activity in a whole vertebrate brain🧠 at single-cell resolution!🐟 70k+ neurons, 3 TB of data & extensive baselines. Connectome is coming! 🧵👇

🎉 Excited to share that we've just hit a remarkable $76M funding milestone, with a $41M round led by @Cathayinnov. We’re grateful to our incredible team, investors, and partners who believe in our mission to build the world's first universal AI foundation model for biology. 🚀

📢 Excited to present our work at #NeurIPS2024! 📄 "Shaving Weights with Occam's Razor: Bayesian Sparsification for Neural Networks using the Marginal Likelihood" Read it here: arxiv.org/abs/2402.15978 Details in 🧵

1/7 Still using Adam? If anyone wants to try a distributed PyTorch implementation of SOAP/eigenvalue-corrected Shampoo with support for low precision data types instead, here you go. github.com/facebookresear…

🌟New work: Noether's razor⭐️ Our NeurIPS 2024 paper connects ML symmetries to conserved quantities through a seminal result in mathematical physics: Noether's theorem. We can learn neural network symmetries from data by learning associated conservation laws. Learn more👇. 1/16🧵

I'm on my way to Vienna to attend #AABI and #ICML2024! Ping me if you want to have a chat :) 🧐 I also have an open #PhDPosition in the area of #causality/causal machine learning. Please reach out via DM/email if you think that you could be a good fit! 🔗rc-trust.ai/groups/causali…

🚨 laplace-torch v0.2 has been released! It’s all about foundation models! You can now easily do probabilistic inference on Huggingface’s LLMs and reward models 🤩 github.com/aleximmer/Lapl… New features! 👇 (1/6) Great effort by @runame_ , @a1mmer, and contributors!

The @BayesDuality workshop starts tomorrow. Full schedule (1 page) bayesduality.github.io/bdw24pictures/… (webpage version with detailed abstract etc.) bayesduality.github.io/talks_2024.html Register at the link in the tweet below to join!

It is my pleasure to announce the 2nd Bayes-Duality workshop, focusing on the design of AI that learns adaptively, robustly, and coninually, like humans. bayesduality.github.io/workshop_2024.… You can watch all 26 talks through zoom livestream (June 12-21). Register at …c59ed978213830355fc8978.doorkeeper.jp/events/172217

Come to our poster session this Tuesday to learn about taming infinitely deep models! #iclr2024

If you're coming to @eccvconf consider our freshly accepted tutorial on: A Bayesian Odyssey in Uncertainty: from Theoretical Foundations 📝 to Real-World Applications 🚀 w/ the amazing @GianniFranchi10 @_olivierlaurent @a1mmer @Pavel_Izmailov More info coming soon #eccv2024

I'm thrilled to share that I'll be joining @RCTrustworthy at @TU_Dortmund to start a research group on Causality this summer! 📢Looking for #PhD students (link attached)! Feel free to reach out via email or dm #PhDposition #Causality #MachineLearning

@JKugelgen @jmhernandez233 @RavikumarPrad Congrats!

On Friday, I successfully defended my PhD thesis "Identifiable Causal Representation Learning: Unsupervised, Multi-View, and Multi-Environment" in Cambridge, examined by Profs @jmhernandez233 and @RavikumarPrad. I'm particularly delighted to have passed with no corrections 🎉

Late announcement, but I'm excited to announce that my PhD thesis is publicly available! In it I give my thoughts on how to approach variational inference in Bayesian neural nets, deep GPs, including discussion on the marginal likelihood and symmetries. arxiv.org/abs/2401.12418

⭐️New paper ⭐️ Excited to share 'The LLM Surgeon', accepted at ICLR 2024. We obtain SOTA pruning performance and even demonstrate structured LLM pruning of full rows and cols. Direct practical impact enabling compression up to 20-30% with negligible loss in performance.🧵1/9👇

Mathematics of deep learning is very challenging, but ultimately fruitful! Our theory paper on "avoiding gradient explosion in networks with batch normalization" was accepted to #iclr !