iiinova — where your digital self keeps evolving

NON-TOXICNo ads · No shady algorithms

iiinova is part of the Atmosphere — an open, public network where you own your identity.

Vincent Fortuin

@vincefort.eurosky.social

6.9K followers 646 following 176 posts

Prof at TU Nuremberg, PI at Helmholtz AI, Fellow at Zuse School for reliable AI, Branco Weiss Fellow, ELLIS Scholar. Prev: TUM, Cambridge CBL, St John's College, ETH Zürich, Google Brain, Microsoft Research, Disney Research. fortuin.github.io

Posts Replies Media

Vincent Fortuin @vincefort.eurosky.social · 05/05/2026

Finally, Geoff Pleiss is arguing that being "good enough" is an achievable goal in Bayesian optimization

030 3

Vincent Fortuin @vincefort.eurosky.social · 05/05/2026

And Chris Oates is now talking about model misspecification in inverse problems

110 1

Vincent Fortuin @vincefort.eurosky.social · 05/05/2026

Now, Clémentine Chazal is telling us about entropy-regularized variational objectives

130 3

Vincent Fortuin @vincefort.eurosky.social · 05/05/2026

@pierrealquier.bsky.social is kicking it off with a talk on PAC-Bayes bounds for Markov chains

151 5

Vincent Fortuin @vincefort.eurosky.social · 05/05/2026

If you're at #AISTATS and not using the day for tourism, come join our workshop OPTIMAL in Medina 1

162 6

Vincent Fortuin @vincefort.eurosky.social · 29/04/2025

I’m greatly enjoying the setting of this year’s #AABI symposium, thanks to NTU Singapore for hosting us!

1121 12

Vincent Fortuin @vincefort.eurosky.social · 25/12/2024

Merry Bayesmas to those who are celebrating 🎅🏻

3491 49

Vincent Fortuin @vincefort.eurosky.social · 12/12/2024

This was a really awesome collaboration with Tristan Cinquin, Marvin Pförtner, @philipphennig.bsky.social, and Robert Bamler!

060 6

Vincent Fortuin @vincefort.eurosky.social · 12/12/2024

🌊 We show that this can improve performance over standard Laplace with weight-space priors in real-world scientific tasks, such as this ocean current modeling problem

120 2

Vincent Fortuin @vincefort.eurosky.social · 12/12/2024

💡 In our work, we propose to use the Laplace approximation in function space! This is mathematically principled (after a bit of measure theory) and can be efficiently implemented using matrix-free linear algebra 🚀

120 2

Vincent Fortuin @vincefort.eurosky.social · 10/12/2024

Very honored to have been elected an @ellis.eu Scholar in the program for Robust Machine Learning: ellis.eu/programs/rob... What better way to celebrate it than to speak at the ELLIS UnConference in Bad Teinach?

2311 31

Vincent Fortuin @vincefort.eurosky.social · 06/12/2024

This was a super fun collaboration, led by my master's student Rayen Dhahri with Alex Immer, @bertrand-sharp.bsky.social, and Stephan Günnemann

031 3

Vincent Fortuin @vincefort.eurosky.social · 06/12/2024

🤔 How efficient can sparsification be without sacrificing performance? ☝️We showcase significant computational savings while retaining high performance across different sparsity levels: 📈 Achieves up to 20x computational savings with minimal accuracy degradation

130 3

Vincent Fortuin @vincefort.eurosky.social · 06/12/2024

Moreover, we introduce Optimal Posterior Damage (OPD), a robust pruning criterion that: ⚡ Reuses posterior precision to guide efficient sparsification ⚡ Covers structured & unstructured pruning seamlessly ⚡ Scales well across various architectures and sparsity levels

120 2

Vincent Fortuin @vincefort.eurosky.social · 06/12/2024

📢 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 🧵

1324 32