Structure-Preserving Transformers for Sequences of SPD Matrices
This work addresses a domain-specific problem in biomedical signal processing by enabling more accurate sleep stage classification from EEG data, though it is incremental as it adapts existing Transformer architectures to a non-Euclidean context.
The paper tackled the problem of classifying sequences of Symmetric Positive Definite matrices by developing a Transformer-based method that preserves their Riemannian geometry, achieving high performance in automatic sleep staging on EEG data.
In recent years, Transformer-based auto-attention mechanisms have been successfully applied to the analysis of a variety of context-reliant data types, from texts to images and beyond, including data from non-Euclidean geometries. In this paper, we present such a mechanism, designed to classify sequences of Symmetric Positive Definite matrices while preserving their Riemannian geometry throughout the analysis. We apply our method to automatic sleep staging on timeseries of EEG-derived covariance matrices from a standard dataset, obtaining high levels of stage-wise performance.