用时间序列模型对齐跨设备数据，实现无需配对样本的标签迁移。
Cross-device Zero-shot Label Transfer via Alignment of Time Series Foundation Model Embeddings
- 通过时间序列基础模型将不同设备信号映射到共享嵌入空间。
- 在嵌入空间中对抗性对齐源域与目标域分布，实现标签迁移。
- 适用于医疗级设备向消费类可穿戴设备的低成本标签转移。
临床级活动记录仪数据虽有高质量、医学验证的标签，但普及的消费类可穿戴设备（如Apple Watch）缺乏此类标注。手动标注成本高且难以扩展。本文提出一种新框架，无需成对数据即可将源域（如活动记录仪）的有价值标签转移到目标域（如Apple Watch）。不直接处理原始时间序列信号，而是利用时间序列基础模型（TSFMs）将两类数据投影至共享潜在嵌入空间，并构建新方法——对抗性对齐时间序列基础模型嵌入（Adversarial Alignment of TSFM Embeddings），使源域与目标域的嵌入分布在此空间内对齐，从而实现跨设备标签迁移。
原文摘要 · Abstract (English)
High-quality, medically validated labels exist for clinical actigraphy data but not for ubiquitous consumer wearables like the Apple Watch. Manually labeling wearables data is expensive and doesn't scale. This paper offers a novel framework that transfers valuable labels from a source domain (e.g., actigraphy) to a target domain (e.g., Apple Watch) without requiring paired data. Instead of working with raw time-series signals, we project both domains into a shared latent embedding space using time-series foundation models (TSFMs) and develop a new framework to align the cross-device representations. Our method, Adversarial Alignment of TSFM Embeddings forces the distributions of source and target embeddings to align within this space, facilitating label transfer across device type.
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。