Reimagining service delivery in the agentic era with Google Public Sector
cloud.google.com·19h ago
AI Research
arXiv CS.AI·May 25·5 min read
VACE: Learning Geometrically Structured Representations for Time Series Anomaly Detection
TL;DR
Detecting anomalies in multivariate time series (data streams with multiple variables) is difficult because abnormal patterns are rare and unlabeled. VACE (Velocity-Aligned Channel Embeddings) uses self-supervised learning to map normal behavior into a compact, geometrically coherent region in embedding space, then flags deviations using position and velocity scores.
✦ Why It Matters
Engineers can deploy simpler, more interpretable anomaly detection systems that explicitly control embedding geometry rather than relying on complex contrastive heuristics.
Key Takeaways
- Utilize VACE for effective anomaly detection in time series data.
- Leverage self-supervised learning without negative samples for training.
- Combine positional and directional scoring for robust anomaly identification.
- Achieve state-of-the-art results on TSB-AD-M benchmark datasets.
How It Works
VACE employs a channel-aware encoder that focuses on aligning normal trajectories in the embedding space. By using a velocity-consistency objective, it ensures that normal data points are represented smoothly and coherently.
This allows the model to define a clear geometric structure where normal behavior is clustered, making it easier to detect anomalies based on their deviation from this structure.