Anomaly Detection

This page explores the application of recent advancements in machine learning (ML) and reinforcement learning (RL) to anomaly detection, a challenging class imbalance problem. Unlike outlier analysis, which focuses on statistically improbable observations, e.g., noise or errors, anomaly detection targets meaningful deviations from the expected process. In anti-money laundering (AML) studies, for example, fraudulent transactions are anomalies requiring investigation; similarly, network intrusions are anomalies with potentially serious consequences. Our focus is on identifying and analyzing these consequential anomalies.





Forthcoming May 2025