STGAD: Training Time Benchmarks for Anomaly Detection Models

STGAD is a dual-score generative-adversarial framework for anomaly detection in multivariate time series. Xiao Liao published benchmark results on figshare in May 2026, showing training times for 5 epochs across five datasets. The 5.5 KB Excel file likely contains performance metrics for models tested on server monitoring, aerospace telemetry, industrial control, and ECG signals.

Use Cases

• Benchmarking model training efficiency based on reported training time metrics
• Comparing adversarial stability of anomaly detection frameworks across different domains
• Evaluating temporal dependency modeling performance on datasets like SMD, SMAP, MSL, SWaT, and MIT-BIH

Strengths

• Training times are benchmarked across five distinct datasets (SMD, SMAP, MSL, SWaT, MIT-BIH)
• Results are published under a CC-BY-4.0 license, allowing for reuse

Limitations

• Row count is unknown, which may limit suitability assessment
• Column-level documentation is absent; field semantics must be inferred after download

Provenance

Source

figshare

Collection Method

Likely contains experimental results from the STGAD framework evaluation.

Freshness

Last updated 2026-05-21 17:42:18