Abstract
Anomaly detection is an important problem with applications in variousdomains such as fraud detection, pattern recognition or medical diagnosis.Several algorithms have been introduced using classical computing approaches.However, using quantum computing for solving anomaly detection problems in timeseries data is a widely unexplored research field. This paper explores the application of quantum autoencoders to time seriesanomaly detection. We investigate two primary techniques for classifyinganomalies: (1) Analyzing the reconstruction error generated by the quantumautoencoder and (2) latent representation analysis. Our simulated experimentalresults, conducted across various ansaetze, demonstrate that quantumautoencoders consistently outperform classical deep learning-based autoencodersacross multiple datasets. Specifically, quantum autoencoders achieve superioranomaly detection performance while utilizing 60-230 times fewer parameters andrequiring five times fewer training iterations. In addition, we implement ourquantum encoder on real quantum hardware. Our experimental results demonstratethat quantum autoencoders achieve anomaly detection performance on par withtheir simulated counterparts.