Rarity-Gated Context Conditioning for Offline Imitation Learning-Based Maritime Anomaly Detection
Researchers have developed a new method called Rarity-Gated Feature-wise Linear Modulation (RGFiLM) to improve anomaly detection in contexts with imbalanced data distributions. This technique uses a rarity score to control how context influences model decisions, making it more decisive in rare situations and conservative in frequent ones. When applied to maritime anomaly detection using AIS and ERA5 data, RGFiLM demonstrated a superior trade-off between F1 score and false positive rate compared to existing methods. AI
IMPACT This method could lead to more reliable anomaly detection systems in domains with highly imbalanced data, such as maritime surveillance.