Researchers have developed AnySleep, a deep learning system capable of staging sleep from various electroencephalography (EEG) and electrooculography (EOG) data with adjustable temporal resolutions. Trained on over 20,000 overnight recordings, AnySleep demonstrates state-of-the-art performance, even outperforming established baselines at 30-second epochs and showing promise for shorter timescale analysis. Separately, a new framework called SleepBand addresses the challenge of single-source domain generalization for sleep staging by incorporating physiologically structured spectral modeling, which anchors representations to invariant sleep rhythms and improves robustness. AI
IMPACT These advancements in AI-driven sleep staging could accelerate research and clinical diagnosis by automating and improving the accuracy of sleep analysis.
RANK_REASON Two research papers introducing new deep learning models for sleep staging.
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
