Cross-subject EEG decoding promises more training data, but it also exposes neural networks to strong inter-subject distribution shifts. We study whether task supervision and architecture alone can learn subject-aligned representations. We replace a shared EEG encoder with subject-specific encoders followed by a common classifier, and compare this hybrid model with standard EEGNet, AttentionBaseNet, and CTNet baselines with Euclidean Alignment (EA) on three motor-imagery datasets and one motor-execution dataset.
EA improves shared encoders by recentering subject covariances, whereas the hybrid encoder reduces reliance on EA: removing EA has little effect on validation-loss dynamics or latent-space organization, and both hybrid variants consistently outperform non-aligned shared baselines. Subject-specific heads increase class distinctiveness and place each subject close to its own latent manifold while improving within-subject class separation. However, on cross-subject classification, subject-specific heads hinder direct parameter transfer to unseen subjects, motivating quantitative head selection and a brief calibration session.
Although decoding gains depend on the dataset and backbone, our main findings concern the solely use of architecture pressure promotes representation learning and alignment in a direction complementary to domain adaptation methods such as Euclidean Alignment. A per-subject low-rank adapter of only \(2\,C\,r\) parameters recovers the full encoder’s accuracy across five backbones and ranks \(r=1\) to 16, so the per-subject module can be compressed by two to three orders of magnitude. BJL, GS, and RYC thank the São Paulo Research Foundation (FAPESP) for the financial support (grants 22/08920-0, 23/06407-7, and 21/12645-2).
University of São Paulo, São Paulo, Brazil Bruna J. Lopes, Gabriel Schwartz & Raphael Y. de Camargo Inria TAU team, LISN-CNRS, Université Paris-Saclay, Orsay, France Bruna J. Lopes, Sylvain Chevallier & Bruno Aristimunha Federal University of ABC (UFABC), Santo André, Brazil Raphael Y. de Camargo & Bruno Aristimunha Swartz Center for Computational Neuroscience (SCCN), Institute for Neural Computation (INC), University of California San Diego, La Jolla, USA Institut de neuromodulation, GHU Paris, psychiatrie et neurosciences, centre hospitalier Sainte-Anne, Université Paris Cité, pôle hospitalo-universitaire 15, Paris, France Bruno Aristimunha has been associated with Yneuro since November 2025.
The other authors declare no competing interests. Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material.
You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Extract — continue reading at the source.
