The abstract is read at the publisher; the summary is JClub's.
ProlateNet, using prolate spheroidal wave function kernels, enables stable cross-sampling-rate transfer for EEG and ECG and achieves a mean worst macro-F1 drop of at most 0.003 on SEED.
- Why it matters: Physiological models often fail when deployed to recording systems with different sampling rates, and existing continuous-kernel approaches cannot be certified before deployment.
- What they did: ProlateNet was compared across fourteen first-layer designs trained at one rate on three public EEG and ECG datasets, and it provides a weights-only certificate with an error bound fixed before training.
- The result: ProlateNet is the only design stable and accurate on all three datasets, and its certificate ranks transfer on SEED with Spearman {rho} = 0.92, versus 0.68 for energy above the training Nyquist frequency.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.