Synthetic Datasets for DINAMO: Dynamic and INterpretable Anomaly MOnitoring for Large-Scale Particle Physics Experiments

This archive contains 1000 synthetic datasets for benchmarking the DINAMO framework (https://arxiv.org/abs/2501.19237), an automated anomaly detection solution featuring both a generalized EWMA-based statistical method and a transformer encoder-based ML approach for Data Quality Monitoring (DQM) in particle physics experiments.The datasets overview: Size: 1000 datasets in .npz format, each containing 5000 runs with one-dimensional Gaussian-based histograms Labels: each run is labeled as "good" (4500 runs) or "bad" (500 runs) Features: datasets mimic particle physics DQM data with emphasis on dynamic operational conditions: Gradual operational drifts (sinusoidal evolution) Abrupt hardware/software changes Varying event statistics and Poisson uncertainties Systematic detector uncertainties Bad runs contain additional distortions and dead histogram bins These datasets enable systematic evaluation of anomaly detection algorithms in time-dependent settings for the DQM problem. More details can be found in the paper and in the GitHub repository at https://github.com/ArseniiGav/DINAMO/