Deep learning methods for detecting anomalies in videos: theoretical and methodological contributions