Weighted Temporal Decay Loss for Learning Wearable PPG Data with Sparse Clinical Labels

2026-02-02

AI Summary

This study addresses the challenge of unreliable long-term supervision in photoplethysmography (PPG) signal analysis for wearable devices, stemming from sparse clinical labels. To mitigate this issue, the authors propose a weighted temporal decay loss training strategy that dynamically adjusts sample weights based on the time interval between PPG segments and their associated clinical labels. The approach incorporates biomarker-specific decay mechanisms and regularization terms to prevent degenerate solutions. Beyond improving predictive performance, the method yields an interpretable metric for the temporal validity of PPG-derived evidence. Evaluated across 450 subjects and 10 biomarkers, the proposed method achieves an average AUPRC of 0.715, outperforming fine-tuned self-supervised models (0.674) and random forests (0.626), with the linear decay variant demonstrating the most robust performance.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Planning, Routing, and Scheduling: Temporal Planning
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Advances in wearable computing and AI have increased interest in leveraging PPG for health monitoring over the past decade. One of the biggest challenges in developing health algorithms based on such biosignals is the sparsity of clinical labels, which makes biosignals temporally distant from lab draws less reliable for supervision. To address this problem, we introduce a simple training strategy that learns a biomarker-specific decay of sample weight over the time gap between a segment and its ground truth label and uses this weight in the loss with a regularizer to prevent trivial solutions. On smartwatch PPG from 450 participants across 10 biomarkers, the approach improves over baselines. In the subject-wise setting, the proposed approach averages 0.715 AUPRC, compared to 0.674 for a fine-tuned self-supervised baseline and 0.626 for a feature-based Random Forest. A comparison of four decay families shows that a simple linear decay function is most robust on average. Beyond accuracy, the learned decay rates summarize how quickly each biomarker's PPG evidence becomes stale, providing an interpretable view of temporal sensitivity.

Problem

• PPG
• sparse clinical labels
• temporal decay
• wearable sensing
• biomarker prediction

Innovation

• Weighted Temporal Decay Loss
• PPG
• Sparse Clinical Labels
• Temporal Sensitivity
• Wearable Health Monitoring