ChatPPG.
Join free
Foundation model

JETS (Empirical Health; joint-embedding FM for irregular behavioural wearable time series)

SpO₂HRActivitysleepwearable-derived metricsNeurIPS 2025 Workshop TS4HWeights: not released
Paper Cite
Venue listedCatalog paper date: 2025Added to atlas: Oct 26, 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.

Hypertension 87%, atrial flutter 70%, ME/CFS 81%, sick sinus syndrome 87%. Beats MAE and PrimeNet baselines. Proprietary data.

Signals
SpO₂, HR, Activity, sleep, wearable-derived metrics
Form factor
Wrist
Year
2025
Architecture

JEPA over (time, value, metric-type) triplet tokens (d=64); context encoder sees a random ~30% of tokens, predictor regresses EMA target-encoder embeddings; Mamba and Transformer encoder variants (per Empirical Health blog)

Pretraining data

Proprietary Empirical Health wearable dataset, ~3M person-days, 63 metric channels (blog)

At a glance
Type
Foundation model
Modalities
SpO₂HRActivitysleepwearable-derived metrics
Form factor
Wrist
Parameters
Not reported
Architecture
JEPA over (time, value, metric-type) triplet tokens (d=64); context encoder sees a random ~30% of tokens, predictor regresses EMA target-encoder embeddings; Mamba and Transformer encoder variants (per Empirical Health blog)
Venue
NeurIPS 2025 Workshop TS4H (poster)
Year
2025
arXiv
Not reported
DOI
Not reported
Code
Not reported
Weights
Not released
Licence
Not reported
Caveats
Not reported
Datasets
Not reported
Papers
Added
Oct 26, 2026
Updated
Oct 26, 2026
JETS (Empirical Health; joint-embedding FM for irregular behavioural wearable time series): ChatPPG research summary card

How it works

How JETS (Empirical Health; joint-embedding FM for irregular behavioural wearable time series) works: Input signals → Foundation model → Downstream tasksInput signals: SpO₂ (Red / IR oximetry); HR; Activity; sleep. Pretraining data: Proprietary Empirical Health wearable dataset, ~3M person-days, 63 metric channels (blog). Foundation model: JETS (Empirical Health; joint-embedding FM for irregular behavioural wearable time series) (State-space model · Transformer).Input signalsSpO₂: Red / IR oximetryR / IRSpO₂Red / IR oximetryHRHRActivityActivitysleepsleepPretraining data: Proprietary Empirical Health wearable dataset, ~3M person-days, 63 metric channels (blog)Pretraining dataProprietary Empirical Health wearable…Foundation modelJETS (Empirical Health…State-space model · Transformer
Pretraining data
Proprietary Empirical Health wearable dataset, ~3M person-days, 63 metric channels (blog)
Architecture
JEPA over (time, value, metric-type) triplet tokens (d=64); context encoder sees a random ~30% of tokens, predictor regresses EMA target-encoder embeddings; Mamba and Transformer encoder variants (per Empirical Health blog)
Inputs
SpO₂, HR, Activity, sleep, wearable-derived metrics
Approach
JEPA · Mamba · irregular time series · behavioral data

Datasets

No public datasets linked yet.

Papers· 1

Cite

APA

Xie, E., Chang, W., Martinez, R. R., & Ballinger, B. (2025). JETS: A Self-Supervised Joint Embedding Time Series Foundation Model for Behavioral Data in Healthcare. NeurIPS 2025 Workshop TS4H. https://openreview.net/forum?id=QqQDjLgHab

BibTeX

@inproceedings{xie2025jets,
  title         = {{JETS: A Self-Supervised Joint Embedding Time Series Foundation Model for Behavioral Data in Healthcare}},
  author        = {Xie, Erik and Chang, Wyatt and Martinez, Raquel Rodriguez and Ballinger, Brandon},
  year          = {2025},
  booktitle     = {NeurIPS 2025 Workshop TS4H},
  url           = {https://openreview.net/forum?id=QqQDjLgHab}
}

Built from the hub record. Check it against the publisher before submitting.

History

  1. Added to the ChatPPG research hub

Connections

Open in the graph explorer
Signal: HR (uses signal)HRHRSignal: SpO₂ (uses signal)SpO₂SpO₂Signal: Activity (uses signal)ActivityActivityPaper: JETS: A Self-Supervised Joint Embedding Time Series Foundation Model for Behavioral Data in Healthcare (paper and model)JETS: A Self-Superv…JETS: A Se…JETS (Empirical Health; joint…JETS (Empirical H…
Signals 3
HR, SpO₂, Activity

Same tasks, data and authors

Papers

Models