Interpretable EEG biomarkers with bag-of-waves: Spatial and temporal waveform dictionaries for low-data regimes
Merged summary
TL;DR - Bag-of-waves learns interpretable dictionaries of recurring EEG waveforms and uses their frequencies and transitions for classification or clustering. It matches strong deep and foundation-model baselines across three datasets while remaining compact and effective with limited data.
- Uses unsupervised shift-invariant k-means to convert continuous EEG into inspectable waveform “atom” tokens.
- Captures temporal structure with atom n-grams and spatial structure with regional and cross-channel atoms.
- Evaluated on mouse genotype clustering, dementia classification, and six-way clinical event classification.
- Recovers known clinical waveform morphologies that neurophysiologists can directly validate.
Sources (1)
Interpretable EEG biomarkers with bag-of-waves: Spatial and temporal waveform dictionaries for low-data regimes
TL;DR - Bag-of-waves learns interpretable dictionaries of recurring EEG waveforms and uses their frequencies and transitions for classification or clustering. It matches strong deep and foundation-model baselines across three datasets while remaining compact and effective with limited data.
- Uses unsupervised shift-invariant k-means to convert continuous EEG into inspectable waveform “atom” tokens.
- Captures temporal structure with atom n-grams and spatial structure with regional and cross-channel atoms.
- Evaluated on mouse genotype clustering, dementia classification, and six-way clinical event classification.
- Recovers known clinical waveform morphologies that neurophysiologists can directly validate.