Learning aligned EEG representations with subject-specific encoders

2026-06-15

AI Summary

This study addresses the significant inter-subject distribution shift that hinders generalization in cross-subject EEG decoding. To mitigate this challenge, the authors propose a hybrid architecture that assigns a dedicated encoder to each subject while sharing a common classifier across subjects. The subject-specific encoders act as learnable alignment mechanisms, effectively internalizing the functionality of traditional Euclidean alignment (EA) and thereby reducing reliance on external alignment procedures. Implemented atop backbone networks including EEGNet, AttentionBaseNet, and CTNet, the approach is validated on four motor imagery datasets. Results demonstrate consistent improvements in classification accuracy and class separability for most subjects, while aligning individual feature representations closer to their intrinsic latent manifolds.

Abstract

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 four motor-imagery datasets. EA improves shared encoders by recentering subject covariances, but the hybrid encoder largely internalises this role: validation-loss curves and latent-distance analyses change little when EA is removed. Subject-specific heads increase class distinctiveness and place each subject close to its own latent manifold, improving most subjects while leaving a method-sensitive subset. These results support subject-specific encoders as a learned alignment mechanism for EEG decoding and identify head selection for unseen subjects as the remaining bottleneck.

Problem

• cross-subject EEG decoding
• inter-subject distribution shifts
• subject-aligned representations
• EEG representation learning
• motor imagery

Innovation

• subject-specific encoders
• cross-subject EEG decoding
• representation alignment
• hybrid neural architecture
• motor imagery