时间序列基础模型可高效适配异常检测，参数高效微调效果优于全量训练。
A Comparative Study of Adaptation Strategies for Time Series Foundation Models in Anomaly Detection
- 对比零样本推理、全模型微调与参数高效微调三种策略。
- 参数高效微调在多数场景下达到或超过全量微调性能，提升效率。
- 适用于数据不平衡场景，适合工业级时序异常检测应用。
时间序列异常检测对复杂系统可靠运行至关重要，但现有方法多需大量任务定制训练。本文探究时间序列基础模型（TSFMs）是否可作为异常检测的通用骨干网络，该模型在大规模异构数据上预训练。通过在多个基准测试上的系统实验，比较了零样本推理、全模型适配和参数高效微调（PEFT）策略。结果表明，TSFMs优于任务特定基线，在严重类别不平衡条件下显著提升AUC-PR与VUS-PR指标。此外，如LoRA、OFT、HRA等参数高效微调方法不仅降低计算成本，且在多数情况下表现匹配或超越全量微调，说明即使以预测为预训练目标，TSFMs亦可高效适配异常检测。这些发现使TSFMs成为可扩展、高效的通用时序异常检测候选方案。
原文摘要 · Abstract (English)
Time series anomaly detection is essential for the reliable operation of complex systems, but most existing methods require extensive task-specific training. We explore whether time series foundation models (TSFMs), pretrained on large heterogeneous data, can serve as universal backbones for anomaly detection. Through systematic experiments across multiple benchmarks, we compare zero-shot inference, full model adaptation, and parameter-efficient fine-tuning (PEFT) strategies. Our results demonstrate that TSFMs outperform task-specific baselines, achieving notable gains in AUC-PR and VUS-PR, particularly under severe class imbalance. Moreover, PEFT methods such as LoRA, OFT, and HRA not only reduce computational cost but also match or surpass full fine-tuning in most cases, indicating that TSFMs can be efficiently adapted for anomaly detection, even when pretrained for forecasting. These findings position TSFMs as promising general-purpose models for scalable and efficient time series anomaly detection.
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。