NV-Tesseract-AD: Curriculum-Enhanced Conditional Diffusion for Multivariate Time-Series Anomaly Detection

📄 arXiv: 2610.12754v1 📥 PDF

作者: Aditi Gautam, Manikandan Ravikiran, Saira Qureshi, Sena Fatma Ekiz, Charan Jagwani, Mahmoud Elhadidy, Sanjeev Roka

分类: cs.LG

发布日期: 2026-10-08


💡 一句话要点

提出NV-Tesseract-AD以解决多变量时间序列异常检测问题

🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture) 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 异常检测 多变量时间序列 条件扩散 课程学习 自监督学习 Transformer 性能提升

📋 核心要点

  1. 现有的多变量时间序列异常检测方法在处理重建难度和噪声水平时存在不足,导致优化不稳定和可区分性差。
  2. 论文提出了一种基于课程学习的条件扩散框架NV-Tesseract-AD,通过结构化的掩蔽和噪声水平训练,提升异常检测的效果。
  3. 实验结果表明,NV-Tesseract-AD在多个基准数据集上实现了显著的性能提升,F1值达到0.984,MSE减少56%。

📝 摘要(中文)

我们提出了NV-Tesseract-AD,一个增强课程学习的条件扩散框架,用于多变量时间序列异常检测。我们的核心假设是,重建难度可以沿着掩蔽严重性和扩散噪声水平两个互补轴线明确构建,同时从简单到困难的任务训练可以稳定优化并提高残差可区分性。我们将基于Transformer的扩散骨干与掩蔽比例和扩散时间步的三阶段双课程相结合,在整个训练过程中保留异常样本,并依赖结构化腐蚀来推动自监督表示学习。我们在TSB-UAD-M语料库上进行分布内训练,并严格零-shot转移到十个EasyTSAD基准系列,未进行每个数据集的调优。双轴课程产生了掩蔽不变的重建,确认了第一训练阶段内互补掩蔽策略之间几乎为零的损失差距,并相较于非课程基线实现了约3.7倍的F1提升和56%的均方误差减少。零-shot性能在NAB上达到F1 = 0.984,在NormA上达到F1 = 0.948。对异构系列的较弱结果追溯到无监督阈值校准,而非重建质量,识别出校准是主要的检测瓶颈。通过DPM-Solver++加速推理将采样成本降低了3.5到5.7倍,但与阈值行为相关联,表明采样器和阈值必须联合验证。这些发现表明,将扩散去噪难度与课程学习原则对齐为复杂多变量时间序列中的稳健异常检测提供了一种有原则且实用的方法。

🔬 方法详解

问题定义:本论文旨在解决多变量时间序列中的异常检测问题,现有方法在处理重建难度和噪声水平时存在优化不稳定和可区分性差的痛点。

核心思路:论文提出的NV-Tesseract-AD框架通过课程学习的方式,将重建任务的难度沿掩蔽严重性和扩散噪声水平两个轴线进行结构化,旨在从简单到困难的训练过程提高模型的稳定性和性能。

技术框架:整体架构包括一个基于Transformer的扩散骨干,结合三阶段的双课程设计,分别针对掩蔽比例和扩散时间步进行训练,同时保留异常样本以推动自监督学习。

关键创新:最重要的技术创新在于双轴课程学习的引入,使得模型在不同的掩蔽策略下保持重建不变性,从而显著提升了异常检测的效果。

关键设计:在参数设置上,采用了结构化的腐蚀方法来驱动自监督表示学习,损失函数设计上强调了重建质量与阈值校准的结合,以应对检测瓶颈。实验中采用DPM-Solver++加速推理,显著降低了采样成本。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果显示,NV-Tesseract-AD在多个基准数据集上表现优异,F1值在NAB上达到0.984,NormA上达到0.948,相较于非课程基线,F1提升约3.7倍,均方误差减少56%。这些结果验证了双轴课程学习的有效性。

🎯 应用场景

该研究的潜在应用领域包括金融欺诈检测、工业设备故障监测和网络安全等。通过提高多变量时间序列异常检测的准确性和效率,NV-Tesseract-AD能够为相关行业提供更为可靠的决策支持,未来可能推动智能监控和预警系统的发展。

📄 摘要(原文)

We propose NV-Tesseract-AD, a curriculum learning-enhanced conditional diffusion framework for multivariate time-series anomaly detection. Our central hypothesis is that reconstruction difficulty can be explicitly structured along two complementary axes, masking severity and diffusion noise level, and that training from easier to harder tasks along both axes simultaneously stabilizes optimization and improves residual discriminability. We integrate a Transformer-based diffusion backbone with a three-stage dual curriculum over masking ratios and diffusion timesteps, retaining anomalous samples throughout training and relying on structured corruption to drive self-supervised representation learning. We evaluate NV-Tesseract-AD under two regimes: in-distribution training on the TSB-UAD-M corpus and strict zero-shot transfer to ten EasyTSAD benchmark families without per-dataset tuning. The dual-axis curriculum produces masking-invariant reconstruction, confirmed by a near-zero loss gap between complementary masking strategies within the first training stage, and yields an approximately 3.7-fold improvement in F1 and a 56 percent reduction in MSE relative to a non-curriculum baseline. Zero-shot performance reaches F1 = 0.984 on NAB and F1 = 0.948 on NormA. Weaker results on heterogeneous families are traced to unsupervised threshold calibration rather than reconstruction quality, identifying calibration as the dominant remaining detection bottleneck. Accelerated inference via DPM-Solver++ reduces sampling cost by a factor of 3.5 to 5.7 at the dataset level but couples to thresholding behavior, establishing that the sampler and threshold must be validated jointly. Together, these findings demonstrate that aligning diffusion denoising difficulty with curriculum learning principles provides a principled and practical approach for robust anomaly detection in complex multivariate time series.