DARE-EEG: A Foundation Model for Mining Dual-Aligned Representation of EEG

2026-05-18

AI Summary

This work addresses the challenge that existing EEG foundation models struggle to learn consistent and transferable representations when masked views exhibit minimal overlap. To overcome this, the authors propose a self-supervised foundation model that explicitly enforces mask invariance during pretraining through dual-alignment representation learning—integrating both mask alignment and anchor alignment mechanisms. Furthermore, they introduce a convolutional-linear probing strategy that enables parameter-efficient adaptation to heterogeneous electrode configurations and sampling rates. The approach synergistically combines contrastive learning, momentum updates, masked signal reconstruction, and decoupled spectral-spatial projection. Evaluated across multiple EEG benchmarks, the method achieves state-of-the-art accuracy while maintaining low parameter complexity and demonstrating strong cross-dataset transferability.

Abstract

Foundation models pre-trained through masked reconstruction on large-scale EEG data have emerged as a promising paradigm for learning generalizable neural representations across diverse brain-computer interface applications. However, a critical yet overlooked challenge is that EEG encoders must learn representations invariant to incomplete observations-when different masked views of the same signal have minimal overlap, existing methods fail to constrain them to a consistent latent subspace, leading to degraded transferability. To address this, we propose DARE-EEG, a self-supervised foundation model that explicitly enforces the mask-invariance property through dual-aligned representation learning during pre-training. Specifically, we introduce mask alignment that constrains representations from multiple masked views of the same EEG sample via contrastive learning, complementing anchor alignment that aligns masked representations to momentum-updated complete features for semantic stability. Additionally, we propose conv-linear-probing, a parameter-efficient strategy that adapts pre-trained representations to heterogeneous electrode configurations and sampling rates through decoupled spectro-spatial projections. Extensive experiments across diverse EEG benchmarks demonstrate that DARE-EEG consistently achieves state-of-the-art in accuracy performance while maintaining relatively low parameter complexity and superior cross-dataset portability compared to existing methods. Furthermore, DARE-EEG contributes to effectively discovering and utilizing the rich potential representations in EEG.

Problem

• EEG
• mask invariance
• incomplete observations
• representation learning
• foundation model

Innovation

• mask-invariance
• dual-aligned representation
• contrastive learning
• conv-linear-probing
• foundation model