A rigorous re-evaluation shows that a simple linear PCA baseline matches or outperforms SOTA Deep Learning models for multivariate time series anomaly detection.
March 20, 2026
Original Paper
Revisiting OmniAnomaly for Anomaly Detection: performance metrics and comparison with PCA-based models
arXiv · 2603.18985
The Takeaway
It exposes a massive evaluation gap in the MTSAD field, showing that complex architectures like OmniAnomaly may not offer real-world advantages over classical methods when benchmarks are controlled for thresholding and point-adjustment biases.
From the abstract
Deep learning models have become the dominant approach for multivariate time series anomaly detection (MTSAD), often reporting substantial performance improvements over classical statistical methods. However, these gains are frequently evaluated under heterogeneous thresholding strategies and evaluation protocols, making fair comparisons difficult. This work revisits OmniAnomaly, a widely used stochastic recurrent model for MTSAD, and systematically compares it with a simple linear baseline base