Institutional Repository of Vilnius University
Quantum autoencoders for anomaly detection /
Abstract
dc:descriptionThis thesis investigates quantum autoencoders for anomaly detection under strong dimension‐ ality constraints. We show that simple QAE architectures suffer from overfitting and poor class sep‐ arability when aggressive compression is applied, leading to elevated false‐positive rates (FPR). To address these limitations, we introduce a BrainBox layer that enhances encoder expressivity while preserving a compact circuit structure. When combined with dropout and ℓ2 regularization, the proposed approach yields more stable latent representations, improves reconstruction fidelity, and reduces FPR while maintaining strong detection performance. Experiments using classical and quantum kNN classifiers demonstrate consistent improvements in F1‐score and PR‐AUC, highlighting the importance of architectural design and regularization in practical quantum anomaly detection.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Yntykbay, Danial,
Subjects
dc:subject × 1Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- eng
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD277678541&prefLang=en_US