LAtte: Hyperbolic Lorentz Attention for Cross-Subject EEG Classification

Johannes Burchert, Ahmad Bdeir, Tom Hanika, Lars Schmidt-Thieme, Niels Landwehr

Keywords: EEG Classification, Hyperbolic Lorentz Attention, Cross-Subject, Low-Rank Adapter

TL;DR: We apply Hyperbolic Lorentz Attention for EEG classification with the utilization of low-rank adapters for subject information embedding.

Abstract: Electroencephalogram (EEG) classification is critical for applications ranging from medical diagnostics to brain-computer interfaces, yet it remains challenging due to the inherently low signal-to-noise ratio (SNR) and high inter-instance variability. To address these issues, we propose LAtte, a novel framework that integrates a Lorentz Attention Module with an InceptionTime-based encoder to enable robust and generalizable EEG classification. Unlike prior work which focused primarily on single-subject performance, LAtte focuses on cross-subject generalization. First, we learn a shared baseline signal across all subjects using pretraining tasks to capture common underlying patterns. Then, we utilize low-rank adapters to learn subject-specific embeddings that model individual differences. This allows a single model to be trained concurrently for all subjects without maintaining separate sets of weights for each individual. We evaluate LAtte on three widely-used EEG benchmarks, achieving a substantial improvement in performance over current state-of-the-art methods.

Primary Area: learning on time series and dynamical systems

Submission Number: 17744