Communication Dans Un Congrès Année : 2026

Balanced Latent Semantics and Signal Fidelity for EEG Representation Learning

Résumé

Electroencephalography (EEG) is critical for neurological diagnosis but suffers from low SNR and subject variability. Current foundation models relying on raw signal reconstruction often overfit to local noise. We propose STELAR, a foundation model with a dual-space objective combining patch-level masked latent prediction for semantic stability with masked reconstruction for raw signal fidelity. To balance these objectives, we introduce MTPE-GB, a validation-driven gradient balancer that adaptively weights tasks without manual tuning or computational overhead. STELAR achieves state-of-the-art linear probing performance across diverse EEG benchmarks, demonstrating robust generalization.

Fichier principal
Vignette du fichier
44_Balanced_Latent_Semantics_a.pdf (1.31 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
licence

Dates et versions

hal-05584741 , version 1 (08-04-2026)

Licence

Identifiants

  • HAL Id : hal-05584741 , version 1

Citer

Van-Chien Nguyen, Trung-Hieu Tran, Tuan-Kiet Doan, Quang Hung Pham, Ngoc-Son Vu, et al.. Balanced Latent Semantics and Signal Fidelity for EEG Representation Learning. ICLR 2026 Workshop on Time Series in the Age of Large Models (TSALM), Apr 2026, Rio de Jaineiro, Brazil. ⟨hal-05584741⟩
83 Consultations
0 Téléchargements

Partager

  • More