A Width-Matched Comparison of Hybrid Quantum-Classical Self-Supervised Learning for Fingerprint Recognition
Hybrid quantum SSL beats matched classical models on fingerprints only for contrastive objectives, not reliably for BYOL.
The paper compares hybrid quantum-classical self-supervised learning for fingerprint recognition by inserting the QuFeX feature extractor into SimCLR, MoCo v2, and BYOL, matched at 8 features equal to 8 qubits. On SOCOFing, with a CIFAR-10 control and k-nearest-neighbor identification, single-run hybrids score higher for both contrastive objectives, while multi-seed BYOL shows no reliable difference. A hardware-efficient QNet circuit does not show the same gain, and the authors discuss architecture, parameter count, nonlinearity, and classical simulability of 8-qubit circuits.
- QuFeX is inserted into SimCLR, MoCo v2, and BYOL at matched 8-feature width.
- Hybrid models beat classical counterparts for contrastive objectives on SOCOFing; BYOL shows no reliable multi-seed gap.
- A hardware-efficient QNet circuit does not reproduce the gain; 8-qubit circuits remain classically simulable.
From the original article
1 min read at arxiv.orgFingerprint 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.
By Maria S. Edwards · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.