Brief

Researchers have developed a new framework called IDEAL (Intelligent Dual dispatch of Emergency AmbuLances) to optimize ambulance dispatching. This system addresses the challenge of dynamic travel times and limited fleet capacity by selectively dispatching a second ambulance only when the predicted travel time difference between primary and secondary routes exceeds a set threshold. IDEAL utilizes a weakly supervised bilevel representation network to learn context-specific travel times from historical data and models uncertainty through Burg-divergence perturbations. The framework was evaluated in collaboration with the Hong Kong Fire Services Department, demonstrating improved response-time and resource trade-offs compared to existing methods. AI

Researchers have developed a new framework called IDEAL for discriminative few-shot anomaly detection. This approach utilizes both normal and anomalous examples as references during inference, unlike previous methods that only used normal references. IDEAL learns intrinsic deviation patterns by first suppressing normal variations and then encoding the remaining deviations into discriminative vectors. This allows the system to generalize to both known and unknown anomalies, outperforming existing methods on eight real-world datasets. AI