用时间衰减权重提升稀疏标签下可穿戴心率数据的健康预测精度
Weighted Temporal Decay Loss for Learning Wearable PPG Data with Sparse Clinical Labels
- 为每类生物标志物设计时间衰减权重，随时间推移降低旧样本权重
- 在450人数据上平均AUPRC达0.715，优于基线模型
- 衰减速率可解释，揭示不同指标的时效敏感性
可穿戴设备与人工智能的发展推动了利用光电容积脉搏波（PPG）进行健康监测的研究。基于此类生物信号开发健康算法的最大挑战之一是临床标签稀疏，导致与实验室检测时间间隔较远的信号难以有效监督。为此，本文提出一种简单训练策略：学习特定生物标志物随时间间隔衰减的样本权重，并在损失函数中引入正则项防止退化解。在来自450名参与者、覆盖10种生物标志物的智能手表PPG数据上，该方法优于基线。在个体级设置下，平均AUPRC为0.715，高于微调的自监督基线（0.674）和基于特征的随机森林（0.626）。四种衰减函数对比表明，线性衰减在多数情况下最稳健。此外，学习到的衰减率能总结各生物标志物的证据时效性，提供可解释的时间敏感性视角。
原文摘要 · Abstract (English)
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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。