Aperiodic and Low-Frequency Spectral Bias in Reconstruction based EEG Foundation Models

• URL: http://arxiv.org/abs/2605.26434v1
• Date: Tue, 26 May 2026 01:40:36 GMT
• Title: Aperiodic and Low-Frequency Spectral Bias in Reconstruction based EEG Foundation Models
• Abstract summary: EEG foundation models, pre-trained on large-scale unlabelled EEG data, have emerged as a promising direction towards learning generalizable EEG representations.<n>We provide a mechanistic account of this shortcoming, attributing it to a fundamental mismatch between reconstruction-based pretext tasks and the idiosyncratic spectral structure of EEG signals.
• Score: 38.88255695601846
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: EEG foundation models, pre-trained on large-scale unlabelled EEG data, have emerged as a promising direction towards learning generalizable EEG representations. Despite showing positive results in data-rich regimes, they often fail to outperform significantly smaller supervised models in low-resource settings compared to fully supervised models. We provide a mechanistic account of this shortcoming, attributing it to a fundamental mismatch between reconstruction-based pretext tasks and the idiosyncratic spectral structure of EEG signals, which decompose into distinct high-power aperiodic and low-power oscillatory components. Using controlled, synthetically-generated EEG inputs, we demonstrate that EEG foundation model embeddings are biased to capture the aperiodic components of the EEG signal while under-representing oscillatory components, particularly at higher frequencies. Additionally, linear probe evaluations on real-world BCI datasets further reveal that embeddings encode subject identity more strongly than task-relevant information, thereby reinforcing the low-frequency and aperiodic component bias in foundation model embeddings trained primarily on reconstruction based objectives. Together, these findings elucidate a failure mode in reconstruction based EEG foundation models and motivate future work to incorporate auxiliary losses explicitly targeting high-frequency oscillatory structure as a path toward more capable and generalizable EEG representations.

Related papers

• Taming foundation model with invariance-oriented pre-training for broad-spectrum EEG analysis across signal-level, brain-state, and brain-health tasks [46.533103231728724]
  INCEPT is an invariance-oriented EEG foundation model trained on over 11,000 hours of unlabelled clinical EEG.<n>We evaluate INCEPT on a broad-spectrum benchmark of ten datasets spanning three levels of post-acquisition EEG analysis.
  arXiv  Detail & Related papers  (2026-08-25T14:20:02Z)
• Understanding and Correcting Low-Frequency Bias in EEG Foundation Model [12.910215709640033]
  We identify a persistent low-frequency bias in representations learned by diverse EEG foundation models.<n>Our analysis links this bias to the interaction between EEG's $1/f$-like spectral structure and neural networks' tendency to preferentially learn low-frequency components.<n>We introduce FAME, a frequency-balanced masked autoencoding framework that reconstructs time--frequency activity in predefined EEG bands from masked EEG inputs.
  arXiv  Detail & Related papers  (2026-08-03T08:36:10Z)
• Let EEG Models Learn EEG [20.64210550961759]
  Just EEG Transformer (Jet) is a generative framework based on conditional flow matching that models EEG as raw sequences evolving along continuous trajectories.<n>Jet consistently achieves state-of-the-art performance, reducing TS-FID by over 40% compared to strong baselines.
  arXiv  Detail & Related papers  (2026-05-20T15:10:10Z)
• Structured Prototype-Guided Adaptation for EEG Foundation Models [56.35233377493694]
  We show that EEG foundation models exhibit poor generalization when subject-level supervision is limited.<n>This failure stems from a structural mismatch between noisy, limited supervision and the highly plastic parameter space of EFMs.<n>We propose SCOPE, a Structured COnfidence-aware Prototype-guided adaptation framework for EFM fine-tuning.
  arXiv  Detail & Related papers  (2026-02-19T10:54:29Z)
• Leveraging Generic Time Series Foundation Models for EEG Classification [8.938023803309097]
  We investigate the applicability of a recently proposed time series classification foundation model, to a different EEG tasks such as motor imagery classification and sleep stage prediction.<n>We find that both variants yield strong performance, consistently outperforming EEGNet, a widely used convolutional baseline, and CBraMod, the most recent EEG-specific foundation model.<n>Our findings highlight the promise of leveraging cross-domain pretrained models for brain signal analysis, suggesting that EEG may benefit from advances in the broader time series literature.
  arXiv  Detail & Related papers  (2025-10-31T14:49:23Z)
• Learning Robust Diffusion Models from Imprecise Supervision [75.53546939251146]
  DMIS is a unified framework for training robust Conditional Diffusion Models from Imprecise Supervision.<n>Our framework is derived from likelihood and decomposes the objective into generative and classification components.<n>Experiments on diverse forms of imprecise supervision, covering tasks covering image generation, weakly supervised learning, and dataset condensation demonstrate that DMIS consistently produces high-quality and class-discriminative samples.
  arXiv  Detail & Related papers  (2025-10-03T14:00:32Z)
• YARE-GAN: Yet Another Resting State EEG-GAN [0.0]
  We implement a Wasserstein GAN with Gradient Penalty (WGAN-GP) to generate multi-channel resting-state EEG data.<n>Our results indicate that the model effectively captures the statistical and spectral characteristics of real EEG data.
  arXiv  Detail & Related papers  (2025-03-04T14:01:10Z)
• Graph Structure Refinement with Energy-based Contrastive Learning [56.957793274727514]
  We introduce an unsupervised method based on a joint of generative training and discriminative training to learn graph structure and representation.<n>We propose an Energy-based Contrastive Learning (ECL) guided Graph Structure Refinement (GSR) framework, denoted as ECL-GSR.<n>ECL-GSR achieves faster training with fewer samples and memories against the leading baseline, highlighting its simplicity and efficiency in downstream tasks.
  arXiv  Detail & Related papers  (2024-12-20T04:05:09Z)
• DGSD: Dynamical Graph Self-Distillation for EEG-Based Auditory Spatial Attention Detection [49.196182908826565]
  Auditory Attention Detection (AAD) aims to detect target speaker from brain signals in a multi-speaker environment. Current approaches primarily rely on traditional convolutional neural network designed for processing Euclidean data like images. This paper proposes a dynamical graph self-distillation (DGSD) approach for AAD, which does not require speech stimuli as input.
  arXiv  Detail & Related papers  (2023-09-07T13:43:46Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.