If you are at #ICML2024 , come chat with Victor Letzelter , David and I about conditional density estimation and ill-posed ML tasks this afternoon ! Poster session 4 - 1:30pm - poster # 1506

To preprocess or not to preprocess your #EEG 🔮when using #ML? 💫An enhanced version of our latest work at Roche is now published in eBioMedicine – The Lancet Discovery Science thelancet.com/journals/ebiom… #DeepDive #Thread 🧵in next post 👉👉🏾👉🏻

. Marie-Constance Corsi started the workshop with an overview of #OpenSource tools, packages, libraries used in #BCI today, for data manipulation, machine learning, etc.

Marie-Constance Corsi Pierre Guetschel presented MOABB, "Mother of all BCI Benchmarks", offering to compare #BCI pipelines with a wide number of existing datasets and state of the art ML algorithms github.com/NeuroTechX/moa…

Marie-Constance Corsi Now it's time for Bruno Aristimunha! Presenting #BrainDecode, bridging the gap between #DeepLearning and #BCI 🧠⚡ github.com/braindecode

why not? hahah

O=

me after learning that diffusion = flow matching

I am happy to announce that this week, I am at NeurIPS Conference! Feel invited to talk with me about open source, EEG, deep learning and brain decoding, or just invite me to drink beer(s) or coffee ;) Ping me on any social media network or whova

I had a great time at NeurIPS Conference! Lots of homework and great to see old and new friends! Looking forward to next year and hoping the jetlag passes soon!

PhD students receiving their stipend.

2024 is 100% complete! Thank you all for following, and have a nice 2025!

Open source is so ungrateful! Now I understand why few people do it for so long

🧵Deep learning simulation of brain activity

No more jobs in the next few hours, AGI delay in a few hours lol

Hell has a special cycle only with slow computers, the most frustrating thing in the world write something and wait seeeeveral seconds until it appears on the screen!

📣 ✨Excited bout our latest work in Patterns, a Cell Press journal Cell Press: GREEN a light-weight #NeuralNetwork with Riemannian Geometry + learnable wavelets for #EEG #biomarkers: doi.org/10.1016/j.patt… Joint work with Joseph & hippneuro ! Thread 🧵👇🏻👇🏾👇

Applying pruning to channel selection, we find up to 89% of channels can be deactivated while maintaining accuracy, for both speech and motor decoders. While this could enable large power savings, the stability of these decoders still needs to be tested online.

I still find myself thinking about that phrase

My super supervisor 🥳🎉