Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

Lecture notes in computer science

Quantum Transfer Learning: The Impact of Classical Preprocessing

Michael Kölle, Jonas Maurer, Philipp J. Altmann, Leo Sünkel, Jonas Stein, Julian Hager, Sebastian Zielinski, Claudia Linnhoff–Popien

Published Jan 1, 2025• 0 citations
Publisher
DOI
10.1007/978-3-031-87327-0_18

Related papers

A Comparative Study of Hybrid Classical–Quantum Transfer Learning

Daniel Martín-Pérez, Francesc Rodríguez-Díaz, Alicia Troncoso +1 more

Communications in computer and information science•Jan 1, 2026•0 citations

Classical–Quantum Transfer Learning for Image Classification

Harshit Mogalapalli, Mahesh Abburi, B. Nithya +1 more

SN Computer Science•Oct 27, 2021•30 citations

Towards Fair Benchmarking of Quantum Transfer Learning for Visual Classification

Nouhaila Innan, Saim Rehman, Muhammad Shafique

Quantum Transfer Learning (QTL) offers a promising approach for visual quantum machine learning under near-term constraints, where limited qubit counts, shallow circuit depths, and costly hybrid optimization restrict end-to-end quantum training. In this setting, pretrained classical backbones can extract high-level visual features, while compact quantum modules operate as trainable classification heads. However, existing QTL results are difficult to compare because they often differ in datasets, preprocessing, backbone settings, qubit budgets, circuit designs, optimization choices, and reporting protocols. This work presents a controlled benchmarking methodology for evaluating representative QTL methods under a unified transfer-learning pipeline. The benchmark compares DQN-QTL, QPIE-QTL, AE-CQTL, PVCQTL, and ED-QTL under shared preprocessing rules, frozen-backbone settings, training conditions, and reporting metrics. The evaluation focuses on Fashion-MNIST and Hymenoptera Ants vs. Bees as the two main datasets, while CIFAR-10 is used to provide additional configuration-level evidence on a harder natural-image task. Beyond predictive performance, the benchmark analyzes circuit size, trainable parameters, quantum parameters, training time, and architectural sensitivity to qubit count and circuit depth. The results show that no single QTL family dominates across all settings: performance depends on the dataset, encoding strategy, circuit design, and computational cost. These findings highlight the need for resource-aware QTL evaluation and provide guidance for selecting hybrid quantum-classical transfer models under near-term resource constraints.

quant-pharXiv:2605.19417•May 20, 2026•0 citations

A Novel Classical-Quantum Transfer Learning Framework for Image Recognition

Banyao Ruan, Zhihao Liu, Xi Li

•Jan 1, 2024•0 citations

Adversarially Robust Quantum Transfer Learning

Amena Khatun, Muhammad Usman

arXiv:2510.16301•Quantum science and technology•Jan 1, 2026•0 citations

On the Quantum Impact in Hybrid Classical-Quantum Transfer Learning

Leo Sünkel, Philipp J. Altmann, Michael Kölle +1 more

As the qubit capacity of current quantum computers is insufficient for many real-world machine learning problems that require the processing of a large number of features, hybrid methods are often used as an alternative for purely quantum models. This includes quantum transfer learning, a hybrid technique that can be applied to a variety of tasks, such as classifying large images. However, as this approach is hybrid in nature, it is not always evident what part of the algorithm is ultimately responsible for the performance. More specifically, while a hybrid method like quantum transfer learning may deliver good results, it is crucial to examine to what extent the quantum part contributed to the overall performance, as this often remains elusive. In this work, we investigate the quantum impact in a hybrid classical-quantum transfer learning approach. We run multiple experiments in various scenarios and show that the impact of the quantum part is, in fact, only minuscule and highly dependent on the classical part of the approach. Our results furthermore indicate that quantum transfer learning does not necessarily provide a significant advantage or improvement over a regular variational quantum circuit approach when the classical part is reduced to a mere feature extractor, and no further classical layers are added to be trained simultaneously to the quantum part.

IEEE International Conference on Quantum Computing and Engineering (QCE)•Sep 15, 2024•1 citation
Quantum Transfer Learning (QTL) offers a promising approach for visual quantum machine learning under near-term constraints, where limited qubit counts, shallow circuit depths, and costly hybrid optimization restrict end-to-end quantum training. In this setting, pretrained classical backbones can extract high-level visual features, while compact quantum modules operate as trainable classification heads. However, existing QTL results are difficult to compare because they often differ in datasets, preprocessing, backbone settings, qubit budgets, circuit designs, optimization choices, and reporting protocols. This work presents a controlled benchmarking methodology for evaluating representative QTL methods under a unified transfer-learning pipeline. The benchmark compares DQN-QTL, QPIE-QTL, AE-CQTL, PVCQTL, and ED-QTL under shared preprocessing rules, frozen-backbone settings, training conditions, and reporting metrics. The evaluation focuses on Fashion-MNIST and Hymenoptera Ants vs. Bees as the two main datasets, while CIFAR-10 is used to provide additional configuration-level evidence on a harder natural-image task. Beyond predictive performance, the benchmark analyzes circuit size, trainable parameters, quantum parameters, training time, and architectural sensitivity to qubit count and circuit depth. The results show that no single QTL family dominates across all settings: performance depends on the dataset, encoding strategy, circuit design, and computational cost. These findings highlight the need for resource-aware QTL evaluation and provide guidance for selecting hybrid quantum-classical transfer models under near-term resource constraints.
As the qubit capacity of current quantum computers is insufficient for many real-world machine learning problems that require the processing of a large number of features, hybrid methods are often used as an alternative for purely quantum models. This includes quantum transfer learning, a hybrid technique that can be applied to a variety of tasks, such as classifying large images. However, as this approach is hybrid in nature, it is not always evident what part of the algorithm is ultimately responsible for the performance. More specifically, while a hybrid method like quantum transfer learning may deliver good results, it is crucial to examine to what extent the quantum part contributed to the overall performance, as this often remains elusive. In this work, we investigate the quantum impact in a hybrid classical-quantum transfer learning approach. We run multiple experiments in various scenarios and show that the impact of the quantum part is, in fact, only minuscule and highly dependent on the classical part of the approach. Our results furthermore indicate that quantum transfer learning does not necessarily provide a significant advantage or improvement over a regular variational quantum circuit approach when the classical part is reduced to a mere feature extractor, and no further classical layers are added to be trained simultaneously to the quantum part.