Abstract
The ongoing deep learning revolution of the last decade has changed how we develop machine learning models in numerous research fields.
However, its adoption in areas like time series has been slower and more constrained compared with images and text domains.
This measured progress underscores the significant challenges in the field.
In the neuroscience application, we face a dual challenge that calls for both clinical neuroscience expertise and machine learning skills to handle complex signals.
Brain decoding aims to extract meaningful information from neural signals to understand brain function or develop applications in healthcare and brain-computer interfaces.
EEG decoding, in particular, presents unique challenges due to the variability across subjects, the limited amount of data, and the need for short calibration times in real-world applications, Chevallier, S. et al. (2024).
Addressing these challenges requires a combination of neuroscience insights and machine learning techniques.
Although deep learning and foundation models hold great promise for these brain decoding problems, current research is in its infancy. Recently, a growing number of works have treated large models to learn common representations across different brain signals.
Despite these advancements, research into EEG decoding still lacks fundamental bases that attempt to compare models fairly and helpfully for real applications.
This special session will explore Machine Learning for Signal Processing in the EEG decoding community at MLSP 2025, examine their clinical applications and interpretable approaches, and discuss the field’s shift toward incorporating more deep learning models.
Call for Papers
Submission Guidelines:
Authors are invited to submit 6 pages full-length papers, including figures and references. All accepted and presented papers will be published in and indexed by IEEE Xplore.
Submissions should follow the official guidelines.
Important Dates
- Paper Submission Deadline for the Special Session: May 10, 2025, 23:59:59 AoE
- Notification of Acceptance: June 24, 2025, 23:59:59 AoE
- Camera-Ready Paper Submission: July 15, 2025, 23:59:59 AoE
Schedule Tentative
| Time | Activity |
|---|---|
| 0:00–0:30 | Introduction and opening remarks - Welcome - Overview of the brain decoding area - Agenda & Logistics |
| 0:30–1:00 | Keynote - Hubert Banville (Meta FAIR - Brain & AI) EEG Decoding in the age of deep learning |
| 1:00–1:30 | Oral Session for the best papers Detailed presentation of the works that had the best review process. |
| 1:30–2:00 | Demos for Session Demonstration of open-source software and clinical cases. |
| 2:00–2:50 | Poster Presentation Poster presentation of the papers. |
| 2:50–3:00 | Closing |
Keynote Speaker - EEG Decoding in the age of deep learning
Hubert Banville is a Research Scientist in the Brain & AI group at Meta FAIR. His research focuses on machine learning for the decoding and processing of functional neuroimaging data. Hubert received his PhD in the Parietal team at Inria, Université Paris-Saclay, where he worked on self-supervised learning for EEG. Previously, he worked on mobile EEG as a researcher at InteraXon (maker of the Muse headband).
Organizers
Our organizing committee benefits from extensive backgrounds, and the research experience of its members spans a broad range of brain signal processing: Electroencephalogram Decoding, Brain-Computer Interface, Neuroscience, Functional connectivity, Deep Learning, and Riemannian Geometry. Additionally, the team comprises academic researchers at different levels of seniority, including one PhD student, a research scientist, an associate professor, and a full professor.
