Model

SMoLK

Cite
Venue listedCatalog paper date: 2023Added to atlas: Nov 6, 2026Source

Activity counts measure attention, not model quality. Missing counts are unknown. Publication status follows the linked source; a listed venue alone does not confirm peer review.

Task model for pointwise PPG artifact segmentation.

Parameters
0.0014M, 0.0085M, 0.0453M
Signals
PPG
Form factor
Wrist
Year
2023
Architecture

Sparse Mixture of Learned Kernels — single-layer multi-scale 1D Conv kernels + weighted sum + sigmoid + optional Savitzky–Golay

At a glance
Type
Task model
Modalities
PPG
Form factor
Wrist
Parameters
0.0014M, 0.0085M, 0.0453M
Architecture
Sparse Mixture of Learned Kernels — single-layer multi-scale 1D Conv kernels + weighted sum + sigmoid + optional Savitzky–Golay
Venue
Not reported
Year
2023
Code
Open
Weights
Not released
Licence
Not reported
Caveats
Not reported
Added
Nov 6, 2026
Updated
Nov 6, 2026

Abstract

Rapid, reliable, and accurate interpretation of medical time-series signals is crucial for high-stakes clinical decision-making. Deep learning methods offered unprecedented performance in medical signal processing but at a cost: they were compute-intensive and lacked interpretability. We propose Sparse Mixture of Learned Kernels (SMoLK), an interpretable architecture for medical time series processing. SMoLK learns a set of lightweight flexible kernels that form a single-layer sparse neural network, providing not only interpretability, but also efficiency, robustness, and generalization to unseen data distributions. We introduce a parameter reduction techniques to reduce the size of SMoLK's networks while maintaining performance. We test SMoLK on two important tasks common to many consumer wearables: photoplethysmography (PPG) artifact detection and atrial fibrillation detection from single-lead electrocardiograms (ECGs). We find that SMoLK matches the performance of models orders of magnitude larger. It is particularly suited for real-time applications using low-power devices, and its interpretability benefits high-stakes situations.

Verbatim fromarXiv· arXiv metadata (abstracts are released under CC0)

SMoLK: ChatPPG research summary card

How it works

How SMoLK works: Input signals → Model → Downstream tasksInput signals: PPG (Photoplethysmogram). Datasets: PPG-DaLiA, WESAD, IEEE Signal Processing Cup 2015 (TROIKA). Model: SMoLK, 0.0014M, 0.0085M, 0.0453M parameters. Downstream tasks: Signal quality; Signal denoising.Input signalsPPG: PhotoplethysmogramPPGPhotoplethysmogramDatasets: PPG-DaLiA, WESAD, IEEE Signal Processing Cup 2015 (TROIKA)DatasetsPPG-DaLiA, WESAD +1 moreModelSMoLK0.0014M, 0.0085M, 0.0453M parametersDownstream tasksSignal qualitySignal qualitySignal denoisingSignal denoising

Datasets linked in the hub; pretraining vs evaluation roles follow the paper.

Architecture
Sparse Mixture of Learned Kernels — single-layer multi-scale 1D Conv kernels + weighted sum + sigmoid + optional Savitzky–Golay
Inputs
PPG
Approach
sqi · motion artifact · edge · ppg only · open code · interpretable

Devices· 2

Full photo credits and licences are on each device page. An icon means we have no openly licensed or manufacturer-cleared photo of that exact device yet.

Datasets used· 3

Cite this model

Cite the publication recorded for this model. Check it against the publisher before submitting.

Chen, S. F., Guo, Z., Ding, C., Hu, X., & Rudin, C. (2023). Sparse learned kernels for interpretable and efficient medical time series processing. Nature Machine Intelligence. https://doi.org/10.1038/s42256-024-00898-4
Chen, Sully F., Zhicheng Guo, Cheng Ding, Xiao Hu, and Cynthia Rudin. 2023. "Sparse learned kernels for interpretable and efficient medical time series processing." Nature Machine Intelligence. https://doi.org/10.1038/s42256-024-00898-4.
Chen, S. F., Guo, Z., Ding, C., Hu, X. and Rudin, C. (2023) 'Sparse learned kernels for interpretable and efficient medical time series processing', Nature Machine Intelligence. doi: 10.1038/s42256-024-00898-4.
Chen, Sully F., et al. "Sparse learned kernels for interpretable and efficient medical time series processing." Nature Machine Intelligence, 2023. https://doi.org/10.1038/s42256-024-00898-4.
S. F. Chen, Z. Guo, C. Ding, X. Hu, and C. Rudin, "Sparse learned kernels for interpretable and efficient medical time series processing," Nature Machine Intelligence, 2023, doi: 10.1038/s42256-024-00898-4.
@article{chen2023sparse,
  title         = {{Sparse learned kernels for interpretable and efficient medical time series processing}},
  author        = {Chen, Sully F. and Guo, Zhicheng and Ding, Cheng and Hu, Xiao and Rudin, Cynthia},
  year          = {2023},
  journal       = {Nature Machine Intelligence},
  doi           = {10.1038/s42256-024-00898-4},
  eprint        = {2307.05385},
  archivePrefix = {arXiv}
}

History

  1. Added to the ChatPPG research hub