A security analyst needs to build a system that flags anomalies in log data without relying on any predefined signatures or human tuning of detection rules. Which deep learning architecture, as taught in the course, is specifically applied to this signature-free log anomaly detection task?