Reproduce “GraphSubDetector: Time Series Subsequence Anomaly Detection via Density-Aware Adaptive Graph Neural Network” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Weiqi Chen, Zhiqiang Zhou, Qingsong Wen, Liang Sun

Researchers at Alibaba Group's DAMO Academy developed GraphSubDetector, a method for time series subsequence anomaly detection that adaptively determines optimal subsequence lengths and leverages a density-aware graph neural network to learn robust representations, achieving superior performance across various real-world datasets.

Key insights

  • Adaptive length selection intelligently identifies optimal temporal contexts, improving the detection of diverse anomalies with varying durations.
  • Density-aware message passing on a learned graph structure theoretically and empirically reduces the variance of normal data and enhances the discriminative power of anomaly representations.
  • A semi-supervised training strategy, incorporating synthetic anomaly injection and a nearest-neighbor contextual anomaly score, effectively addresses the scarcity of labeled anomaly data.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.