BRAIn: Better Representations for Artificial Intelligence
Texte
BRAIN is a research team of Lab-STICC (CNRS UMR 6285) hosted at IMT Atlantique and part of the Mathematical and Electrical Engineering (MEE) department. The purpose of BRAIn is to investigate key questions at the crossbreed of Artificial Intelligence, Deep Learning and Signal Processing, with applications using images, sounds, text and more complex domains including neuroimaging data.
As of 2024, these questions include:
- Thrifty AI, or how to find efficient representations in the low data regime (typically few-shot or few-label learning)
- Compression of AI, or how to find efficient representations to lower the demand in computations and energy of AI systems, typically for AI on edge
- Geometry and AI, or how to leverage the geometry of latent spaces or input domains to better adapt and monitor AI systems
Feats & Distinction:
- Few-shot learning
- New paper at CVPR 2025, ProKeR: A Kernel Perspective on Training-Free Few-Shot Adaptation of Large Vision-Language Models
- New paper published to be published on TMLR Oops, I Sampled It Again on reinterpretation of confidence intervals in few-shot learning
- Our grad students have won the AMD Open Hardware Competition (https://www.openhw.eu/2023-results-gallery) with the PEFSL project (https://github.com/brain-bzh/PEFSL) in 2023
- Compression of Deep Neural Networks
- A particularly visible post of the team on towardsdatascience : https://towardsdatascience.com/neural-network-pruning-101-af816aaea61 with 300+ claps
- Strong industrial partnerships on deployment of neural networks on microcontrollers (GoodFloow, Schneider)
- Neuroimaging and Cognitive Neurosciences
- BRAIn collaborates with Mila, University of Southern California, University of Rochester
- Graph Signal Processing for Neuroimaging and cognitive neurosciences (with CoCoLab – Karim Jerbi)
- Applications to Brain Computer Interfaces and Neurofeedback, with EMPENN Team (INRIA)
- Sound and temporal data
- Discover how the Silent Cities project captured global changes in urban soundscapes during the COVID-19 pandemic, with open-source data and technical validation—read the full article on Nature Scientific Data Silent Cities
- Rank #2 and Jury Award on DCASE 2023 Task 5
