Generative Adversarial Networks With AdaBoost Ensemble Learning for Anomaly Detection in High-Speed Train Automatic Doors

Due to the scarcity of abnormal condition data in components transportation systems, only normal are typically used train models for anomaly detection. One main challenges is difficulty properly representing distribution which non-smooth, high-dimensional and on a manifold. This work develops an detection model based Auto-Encoder (AE) formed by generator Generative Adversarial Network (GAN) auxiliary encoder capture sophisticated structure. The reconstruction error AE is, then, as score detect anomalies. Additionally, adaptive noise added make easier GAN optimization, AdaBoost-based ensemble learning scheme improve performance new approach setting hyperparameters AE-GAN derivation lower bound Jensen-Shannon divergence between distributions developed. method has been applied synthetic real collected from automatic doors high-speed trains.