Mitigating Subject Dependency in EEG Decoding with Subject-Specific Low-Rank Adapters

2025-10-09

AI Summary

To address subject-specific distribution shifts in cross-subject EEG decoding, this paper proposes the Subject-Conditioned Layer (SCL), a plug-and-play conditional layer module. SCL decomposes layer weights into a shared backbone and subject-specific low-rank adapters, explicitly decoupling universal representations from individualized features. It supports end-to-end training and is compatible with both linear and convolutional architectures. Empirical evaluation across multiple EEG benchmark tasks demonstrates that SCL significantly improves cross-subject generalization performance and stability compared to both unified-weight models and ensembles of independently trained subject-specific models. These results validate the effectiveness and practicality of low-rank conditional modeling for foundational neural interface models, offering a scalable and parameter-efficient approach to subject adaptation without sacrificing architectural flexibility or training efficiency.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Computer Vision: Representation Learning for Vision

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Subject-specific distribution shifts represent an important obstacle to the development of foundation models for EEG decoding. To address this, we propose Subject-Conditioned Layer,, an adaptive layer designed as a drop-in replacement for standard linear or convolutional layers in any neural network architecture. Our layer captures subject-specific variability by decomposing its weights into a shared, subject-invariant component and a lightweight, low-rank correction unique to each subject. This explicit separation of general knowledge from personalized adaptation allows existing models to become robust to subject shifts. Empirically, models equipped with our layer outperform both a shared-weight-only model (subject-agnostic model) and the average of individually trained subject-specific models. Consequently, the Subject-Conditioned Layer, offers a practical and scalable path towards building effective cross-subject foundation models for EEG.

Problem

• Mitigating subject dependency in EEG decoding models
• Addressing subject-specific distribution shifts in EEG data
• Building robust cross-subject foundation models for EEG

Innovation

• Uses low-rank adapters for subject-specific EEG decoding
• Decomposes weights into shared and personalized components
• Enables robust cross-subject foundation models for EEG