Resonant Anomaly Detection

Resonant anomaly detection aims to identify unusual patterns in data, particularly those indicative of rare events or "signals" amidst a larger background. Current research emphasizes developing model-agnostic methods, often employing normalizing flows or autoencoders, that can effectively distinguish signals from background noise even with limited labeled data and multiple, potentially heterogeneous, reference datasets. These advancements are improving the sensitivity and interpretability of anomaly detection across diverse fields, from high-energy physics to financial auditing, enabling more robust identification of significant deviations and facilitating deeper understanding of underlying phenomena.