aiqbench analysis
This research proposes a personalized federated learning framework that separates shared neural network trunks from subject-specific heads to improve EEG decoding accuracy without pooling data. The study demonstrates that this architecture outperforms standard federated and centralized training methods in motor-imagery tasks while reducing communication overhead. By addressing inter-subject variability directly, the approach offers a practical path toward scalable, privacy-preserving brain-computer interface models.
Personalised federated learning for Riemannian and Euclidean EEG decoding
arXiv:2609.29037v1 Announce Type: cross Abstract: Federated learning (FL) lets EEG decoders learn from recordings of several subjects without pooling them. We consider two light EEG decoders, the Riemannian SPDNet and the Euclidean EEGNet. Both split into a trunk, which builds a latent representation, and a head, which classifies it. Inter-subject variability, however, makes a single shared FL model a poor fit for each subject. Personalised FL addresses this: all subjects learn a common trunk, and each subject keeps its own head. We adapt it for SPDNet and study its effects against standard FL and centralised training, with EEGNet as a Euclidean baseline. Experiments cover three motor-imagery datasets that span diverse regimes in channels, subjects and classes. We observe that personalised SPDNet reaches higher accuracy than both standard FL and centralised training, while converging in fewer rounds and communicating fewer parameters than standard FL. It also outperforms every EEGNet configuration on two of the three datasets, although centralised EEGNet outperforms centralised SPDNet.
Abstract:Federated learning (FL) lets EEG decoders learn from recordings of several subjects without pooling them. We consider two light EEG decoders, the Riemannian SPDNet and the Euclidean EEGNet. Both split into a trunk, which builds a latent representation, and a head, which classifies it. Inter-subject variability, however, makes a single shared FL model a poor fit for each subject. Personalised FL addresses this: all subjects learn a common trunk, and each subject keeps its own head. We adapt it for SPDNet and study its effects against standard FL and centralised training, with EEGNet as a Euclidean baseline. Experiments cover three motor-imagery datasets that span diverse regimes in channels, subjects and classes. We observe that personalised SPDNet reaches higher accuracy than both standard FL and centralised training, while converging in fewer rounds and communicating fewer parameters than standard FL. It also outperforms every EEGNet configuration on two of the three datasets, although centralised EEGNet outperforms centralised SPDNet.
Machine Learning (stat.ML); Machine Learning (cs.LG)
arXiv:2609.29037 [stat.ML]
(or arXiv:2609.29037v1 [stat.ML] for this version)
https://doi.org/10.48550/arXiv.2609.29037 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 253 words
Related coverage
Intrinsic-Hybrid Latent Diffusion Models for Generative Modeling on Unknown Manifolds
arXiv · Aug 6, 2026
Riemannian Graph Tokenizer for Structural Knowledge Transfer
arXiv · Oct 14, 2026
High-Resolution Dynamic Functional Connectivity Generation with Graph-Variate Flow Matching
arXiv · Sep 30, 2026
Directional Pullback Regularization for Long-Horizon Rollout of Latent World Models in Autonomous Driving
arXiv · Oct 14, 2026