Hybrid quantum classical models improve fingerprint recognition with contrastive learning
A Width-Matched Comparison of Hybrid Quantum-Classical Self-Supervised Learning for Fingerprint Recognition
Machine Learning
Summary
Fingerprint recognition systems usually need lots of labeled examples to learn well. The authors studied whether combining quantum computing parts with classical machine learning can help systems learn from unlabeled data, called self-supervised learning. They tested three types of learning methods and found that the quantum-classical mix helped improve accuracy only in certain learning approaches. They also checked if these improvements were really from the quantum parts or from other factors in the design. This shows the benefit depends on the specific learning method used.
What this means in practice
- •For biometric system developers: Improve fingerprint recognition by integrating quantum feature extraction modules into contrastive self-supervised learning models for better unlabeled data efficiency.
- •For quantum algorithm engineers: Design quantum-classical hybrid circuits focusing on self-supervised contrastive objectives to enhance learning representations in biometric tasks.
Tested on one dataset.
Authors
Maria S. Edwards, Kidwell Dlamini, Pin-An Lin, Wen-Hsien Hsu, Wen-Chieh Fang
Abstract
Fingerprint recognition is a widely deployed biometric, but supervised training requires large labeled enrollment sets. Self-supervised learning (SSL) removes this requirement, and hybrid quantum-classical models have been proposed to enrich the learned representations. Prior quantum SSL studies consider a single contrastive objective, so it is unclear whether reported benefits depend on the objective or can be attributed to the quantum circuit. We insert the QuFeX quantum feature-extraction module into three SSL frameworks, the contrastive SimCLR and MoCo v2 and the non-contrastive BYOL, and compare each hybrid with its classical counterpart at matched representation width (8 features, equal to 8 qubits) on the SOCOFing fingerprint dataset, with a CIFAR-10 control, using k-nearest-neighbor identification on encoder features. In single-run experiments the hybrid scores clearly higher for both contrastive objectives, whereas for BYOL a multi-seed analysis shows no reliable difference, suggesting that any benefit depends on the SSL objective. A hardware-efficient circuit (QNet) does not show the same gain. We examine whether the gains can be attributed to the quantum circuit, considering circuit architecture, trainable parameter count, nonlinearity, and the classical simulability of 8-qubit circuits.