👩‍🎓👨‍🎓 Internship offers (1st step to PhD program) in my group: team.inria.fr/soda/job-offer… Topics: ◼ Health AI & causality, accounting for censoring (for people who love health impact) ◼ Foundation models for tabular learning (for people into bigger models) Come work with us!

Hello #OHBM2024! Wanna know how you can leverage across-subject information to improve within-subject decoding? @BertrandThirion, Liza Al-Shikhley, and I came up with a very approachable method to do so. Let's talk more about it at poster #1432 on Wed and Thu between 2-4pm.

The #IBC 3rd-release data-paper is out! We validate naturalistic tasks using #FastSRM, an unsupervised model where a shared response and individual maps are learnt jointly from fMRI time series. Check it out! doi.org/10.1038/s41597-024-0… @BertrandThirion @hr1ch3rd @NeuroSpin_91

#Cerveau🧠| Première mondiale : le cerveau dévoilé comme jamais grâce à l’IRM le puissant au monde conçu et développé au CEA ! 🤩Découvrez les 1ères images de cerveau obtenues avec le scanner IRM Iseult, doté d’un champ magnétique inégalé de 11,7 teslas👉bit.ly/Première-mondiale-Is…

What drives innovation in machine learning? In a new paper, we argue that application-driven work is systemically under-valued in the machine learning community, but that it's essential for both innovation and impact. arxiv.org/abs/2403.17381 1/3

✨ embedder for short strings, useful for table string entries (string matching, entity linking or clustering, fuzzy join) 🤖 huggingface.co/Lihuchen/pear… Adds to transformer models a character-level model and a model of the phrase type. Yet very lightweight (33.4M or 109M params) 👇

Announcing @probabl_ai, an accelerator for open source and @scikit_learn in particular. It is both a spin-off from @inria and a broader venture. I'm excited and proud to be part of this adventure!

We still need many more native English volunteers for our simple test of math vocabulary -- click on this link: neurospin-data.cea.fr/exp/ma… Please pass the word around! it only takes 15 minutes.

🎉✨First release of skrub 0.1.0 skrub-data.org Couple dataframes and databases to machine learning to facilitate data prep 🌟Less data wrangling, more machine learning🌟 1/8 🧵👇

🚀Happy to share our paper at @NeurIPSConf with @BertrandThirion @ogrisel @pneuvial: arxiv.org/abs/2310.10373 It introduces KOPI, a novel method for statistically controlled variable selection based on Knockoffs. Come check out our poster #1004 at #NeurIPS2023! 🧵1/6

Are you interested in assessing variable importance in your custom non-linear or non-parametric prediction model for high-dimensional biomedical data? Do you prefer methods based on statistical principles rather than ad-hoc rankings of variables? 🚨 #NeurIPS23 📢 Our latest work presented at #NeurIPS23 addresses the challenge of statistically valid inference of variable importance through conditional permutation techniques and presents novel model-agnostic algorithms🚀💥 alongside simulation-based and empirical benchmarks versus popular variable importance techniques. If you are in New Orleans 🎷🥁🎶 - don't miss out on chatting with @ahmadchm96 & @BertrandThirion on Wednesday at poster session 3. Both attend in person. For a detailed thread 🧵 👇🏿👇👇🏽x.com/dngman/status/17335135…

🥁📣Check out our latest paper that might help you combining heterogeneous #EEG datasets for #ML applications!👇🧶

Félicitations @GaelVaroquaux 👏 ! Directeur de recherche et responsable de l'équipe-projet @soda_INRIA, ses travaux récents ont démontré le décalage entre la performance des algorithmes d'IA et la réalité du progrès clinique. #HighlyCitedResearchers ➡️ inria.fr/fr/imagerie-medical…

@huggingface proudly claims to be ethical, yet they exhibit 3 lamas for their event, when activists of animal rights try to ban all exploitation of animals for leisure. I'm curious: how does it fit in your ethical framework?

pyFaust goes open source! gitlab.inria.fr/faustgrp/fau… FAµST factorizes any dense matrix as a product of sparse factors. Apply it to your operators to get reduced storage and faster matrix multiplication!

Conditional independence of entries of Gaussian vectors defines a graph associated to the inverse of the covariance (precision matrix). Graphical Lasso estimates a sparse graph by l^1 penalizing the maximum likelihood estimator. en.wikipedia.org/wiki/Graphi…

🎓👨‍🦱👩 Post-doc: From missing values to deep learning on sets team.inria.fr/soda/job-offer… with myself and @MarineLeMorvan at @soda_INRIA Come work with us on an exciting topic across statistics and deep learning

🏆Most cited paper of 2022 is this review on #MachineLearning for medical imaging, including its methodological failures and recommendations for the future. @GaelVaroquaux @DrVeronikaCH go.nature.com/470KQRL

I am launching @mystralai, a gen AI agency. We develop custom LLM enterprise applications We serve $100M ARR clients and fast growing series A/B startups I am bold on the opportunity with AI service biz right now My first time running an agency. Will share our lesson learned!

Woodbury formula allows one to compute in two different ways the solution of Ridge regression. It allows one to even use infinite-dimensional feature spaces (p=inf) via kernelization. en.wikipedia.org/wiki/Woodbu… en.wikipedia.org/wiki/Tikhon…