![[2602.19138] CRCC: Contrast-Based Robust Cross-Subject and Cross-Site Representation Learning for EEG](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.19138] CRCC: Contrast-Based Robust Cross-Subject and Cross-Site Representation Learning for EEG
Summary
The paper presents CRCC, a novel framework for improving EEG-based neural decoding models' generalization across different acquisition sites, addressing bias factors that hinder performance.
Why It Matters
This research is significant as it tackles the challenge of site-dependent biases in EEG data, which can limit the effectiveness of neural decoding models in clinical settings. By introducing a robust training paradigm, CRCC enhances the reliability of EEG interpretations across diverse environments, potentially improving diagnostic accuracy for conditions like Major Depressive Disorder.
Key Takeaways
- CRCC identifies and mitigates three fundamental bias factors affecting EEG data generalization.
- The framework employs a two-stage training paradigm combining encoder-decoder pretraining and contrastive learning.
- CRCC demonstrates a 10.7 percentage-point improvement in balanced accuracy under zero-shot site transfer.
Quantitative Biology > Neurons and Cognition arXiv:2602.19138 (q-bio) [Submitted on 22 Feb 2026] Title:CRCC: Contrast-Based Robust Cross-Subject and Cross-Site Representation Learning for EEG Authors:Xiaobin Wong, Zhonghua Zhao, Haoran Guo, Zhengyi Liu, Yu Wu, Feng Yan, Zhiren Wang, Sen Song View a PDF of the paper titled CRCC: Contrast-Based Robust Cross-Subject and Cross-Site Representation Learning for EEG, by Xiaobin Wong and Zhonghua Zhao and Haoran Guo and Zhengyi Liu and Yu Wu and Feng Yan and Zhiren Wang and Sen Song View PDF HTML (experimental) Abstract:EEG-based neural decoding models often fail to generalize across acquisition sites due to structured, site-dependent biases implicitly exploited during training. We reformulate cross-site clinical EEG learning as a bias-factorized generalization problem, in which domain shifts arise from multiple interacting sources. We identify three fundamental bias factors and propose a general training framework that mitigates their influence through data standardization and representation-level constraints. We construct a standardized multi-site EEG benchmark for Major Depressive Disorder and introduce CRCC, a two-stage training paradigm combining encoder-decoder pretraining with joint fine-tuning via cross-subject/site contrastive learning and site-adversarial optimization. CRCC consistently outperforms state-of-the-art baselines and achieves a 10.7 percentage-point improvement in balanced accuracy under strict zero-shot site...
