Reproduce “Anomaly Detection at Scale: The Case for Deep Distributional Time Series Models” 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
Fadhel Ayed, Lorenzo Stella, Tim Januschowski, Jan Gasthaus
The authors model time series of interval distributions to detect anomalies that single summary statistics may hide. They represent measurements with histograms and use a recurrent network to forecast each interval’s distribution from past data and optional covariates. On AWS benchmark B1, the method achieves ROC-AUC 0.8183, above TwitterAD and Luminol, but this labeled benchmark does not establish production-wide accuracy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.