PaperarXiv:2611.14849
Generalize Then Personalize: Wearable-Based Heart Rate Forecasting Under Scarce Ruck-March Data
IEEE BIBM 2026 (arXiv preprint)
TL;DR
HR forecasting (30/60 s ahead) during loaded ruck marches by 8 US Army medics. Population pretraining then per-person fine-tuning (PPFT) beat population-only and personal-from-scratch models. A naive persistence baseline still had the best 30 s MAE.
Abstract
Heart rate (HR) forecasting may provide advance information about near-future cardiovascular response, potentially supporting physiological monitoring and workload-management decisions during ruck marching. However, robust model development is constrained by the difficulty of collecting large-scale ruck-march datasets, creating a scarce-data problem in which models must generalize across participants while adapting to individuals from limited personal data. To address this problem, we benchmark population-only training (POT), personalized-from-scratch training (PFST), and population-pretrained fine-tuning (PPFT) for HR forecasting during loaded ruck marching using data from eight active-duty U.S. Army medics across thirteen sessions. LSTM, Attention-LSTM, and CNN-LSTM backbones were evaluated against a persistence baseline at 30 s and 60 s horizons. Across all neural architectures and horizons, PPFT yielded lower mean MAE, RMSE, and endpoint MAE than both POT and PFST. Participant-level paired analysis after averaging across three training seeds showed that, relative to PFST, PPFT reduced trajectory MAE by 2.22-4.30 bpm and endpoint MAE by 1.24-2.88 bpm. All six trajectory-MAE comparisons remained significant after Holm adjustment (p_adj = 0.0469), while four of six endpoint-MAE comparisons remained significant after adjustment. At 30 s, persistence achieved the lowest MAE (3.555 ± 1.274 bpm), while CNN-LSTM PPFT achieved the lowest RMSE (4.669 ± 1.916 bpm); LSTM PPFT achieved the lowest endpoint MAE (5.093 ± 2.365 bpm). At 60 s, Attention-LSTM PPFT achieved the lowest MAE (5.573 ± 1.939 bpm) and RMSE (7.134 ± 2.260 bpm), while CNN-LSTM PPFT achieved the lowest endpoint MAE (7.553 ± 2.357 bpm). These results suggest that population pretraining followed by participant-specific fine-tuning may provide a practical strategy for scarce-data wearable HR forecasting.
Verbatim from arXiv (opens in a new tab) · arXiv metadata (abstracts are released under CC0)
Results on ChatPPG leaderboards
All leaderboardsNo results from this paper on ChatPPG leaderboards yet. Browse the leaderboards.
Tasks & signals
Tasks
Signals
Keywords
How it works
Cite this paper
Raju, M. A., Ma, R., Ramsay, J. W., & Cornell, D. J. (2026). Generalize Then Personalize: Wearable-Based Heart Rate Forecasting Under Scarce Ruck-March Data. IEEE BIBM 2026. https://arxiv.org/abs/2611.14849
Raju, Muntaqim Ahmed, Ruizhe Ma, John W. Ramsay, and David J. Cornell. 2026. "Generalize Then Personalize: Wearable-Based Heart Rate Forecasting Under Scarce Ruck-March Data." IEEE BIBM 2026. https://arxiv.org/abs/2611.14849.
Raju, M. A., Ma, R., Ramsay, J. W. and Cornell, D. J. (2026) 'Generalize Then Personalize: Wearable-Based Heart Rate Forecasting Under Scarce Ruck-March Data', IEEE BIBM 2026. Available at: https://arxiv.org/abs/2611.14849
Raju, Muntaqim Ahmed, et al. "Generalize Then Personalize: Wearable-Based Heart Rate Forecasting Under Scarce Ruck-March Data." IEEE BIBM 2026, 2026. https://arxiv.org/abs/2611.14849.
M. A. Raju, R. Ma, J. W. Ramsay, and D. J. Cornell, "Generalize Then Personalize: Wearable-Based Heart Rate Forecasting Under Scarce Ruck-March Data," IEEE BIBM 2026, 2026, [Online]. Available: https://arxiv.org/abs/2611.14849.
@article{raju2026generalize,
title = {{Generalize Then Personalize: Wearable-Based Heart Rate Forecasting Under Scarce Ruck-March Data}},
author = {Raju, Muntaqim Ahmed and Ma, Ruizhe and Ramsay, John W. and Cornell, David J.},
year = {2026},
journal = {IEEE BIBM 2026},
eprint = {2611.14849},
archivePrefix = {arXiv}
}Added to the ChatPPG research hub