The Workings of Quantum Anomaly Detection(One-class Classifiers)
This repository contains the workings of a research project made at Mastricht University finished on June 2023:
Quantum computing has sparked a global frenzy, captivating minds with its immense potential to transform the world of computation. Among its many promises, anomaly detection stands tall as a domain ripe for disruption. Why, some might ask? The answer lies in this technology’s potential speed and precision, enabling us to identify elusive and extraordinary data points or behaviors. In this project, we explore the current landscape of anomaly detection methods, pitting the quantum realm against its classical counterpart. Our quest involves comparing tried-and-true methodologies, including kernel PCA, support vector machines, and fully connected autoencoders, against their quantum counterparts — the variational quantum classifier as one class classifier and the quantum support vector machine. As our investigation unfolds, a revelation emerges, a noticeable implementation gap separating these methodological categories. The scales tip overwhelmingly in favor of classical methods, their established prowess shining through. However, as we delve deeper into the intricacies of complex and inseparable data, a lingering question emerges: Could quantum anomaly detection techniques hold a hidden advantage for specific data types intertwined with quantum elements?
The full report is available.
In the folders the following files can be founded:
- Data extraction, analysis and preprocessing
- Building of classical (SVM, Kernel PCA and Autoencoder) and quantum models (QSVM and VQC)
- Figures with the performance of the models