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

arXiv:2605.25365

Quantum Parameterized Self-Attention Network for Image Classification

Wenwei Zhang, Jintao Wang, Tianyu Ye, Changgeng Liao

quant-phSubmitted May 25, 2026• Announced May 26, 2026• 1 citation
arXivHTMLPDF

Abstract

Transformer now underpins modern AI as its core infrastructure. Its defining capability-dynamically focusing on the most relevant information in complex inputs-is bounded above by the self-attention scoring function. Quantum computing, with its superposition, entanglement, and probabilistic outputs, offers a fundamentally distinct computational framework for exploring beyond the design constraints of classical scoring functions. While quantum attention mechanisms have shown initial promise, existing works remain largely confined to redefining feature similarity measures, leaving the systematic use of parameterized quantum circuits (PQCs) as scoring functions largely unexplored; a substantial portion of existing schemes further rely on purely quantum architectures, precluding effective encoding of high-dimensional image inputs in the Noisy Intermediate-Scale Quantum era. We propose the Quantum Parameterized Self-Attention Network (QPSAN), implementing the self-attention scoring function via PQCs with only 5 trainable quantum parameters per layer. QPSAN computes query-key attention scores through quantum state encoding and joint measurement, yielding naturally bounded outputs without the explicit scaling of classical dot-product attention. We further establish a theoretical framework of the mathematical properties of this scoring function, demonstrating its potential to capture complex nonlinear query-key interactions, and quantifying the structural constraints of the encoding layer via effective degrees of freedom analysis. Experiments on four vision datasets show that QPSAN significantly outperforms the Vision Transformer (ViT) baseline, with the quantum representational advantage amplifying as data complexity increases. Ablation studies indicate that the performance gains may stem from the structural inductive bias of the quantum circuit rather than from parameter scale.

Comments
16 pages, 7 figures, 7 tables

Related papers

A Quantum Self-Attention Neural Network Model on Quantum Circuits

Jin Yan Zheng, Qing Zhong Gao, Daoyi Dong +2 more

This article proposes a quantum self-attention neural network (QSAN) model that is implementable on parameterized quantum circuits (PQCs), providing a novel avenue for addressing natural language processing (NLP) tasks via quantum computing. The QSAN architecture comprises four blocks: the data preprocessing block, the quantum encoding block, the model design block, and the network optimization block. Through these blocks, classical text data are initially preprocessed and encoded into quantum states. The rich semantic features and intricate relationships within these quantum states are then captured and learned by a quantum self-attention layer and a quantum fully connected layer. Finally, the model performance is enhanced by optimizing the network parameters. Simulation results on publicly available topic classification and sentiment analysis datasets demonstrate that the QSAN model outperforms state-of-the-art baselines. As a limitation of the current study, these evaluations are restricted to a maximum sentence length of eight words due to simulation and hardware constraints. Moreover, the effectiveness, scalability, trainability, and robustness of the proposed model are comprehensively evaluated, highlighting its superior performance and potential for advanced NLP tasks.

IEEE Transactions on Cybernetics•Jun 29, 2026•0 citations

Efficient Circuit Classifier Design for Enhancing Quantum Self-Attention in Vision Transformers

Huyen Quang Tran, Yu-Han Lin, Hanh Thi Minh Tran +2 more

Quantum self-attention is a promising approach to replace traditional self-attention mechanisms in classical deep learning, which often require a massive number of GPUs for training. Although several early studies have proposed quantum self-attention architectures, these models often struggle with classification tasks in vision transformers. In this study, we address this limitation by enhancing quantum self-attention through integration with circuit classifier design and, crucially, by optimizing the number of strong entangling layers. Evaluation results show that incorporating an optimal number of entangling layers substantially improves classification performance, outperforming models that rely solely on directly measured quantum self-attention. We further analyze how varying the number of quantum classifier layers affects classification accuracy during training. Experiments on downscaled MNIST and MC datasets reveal that the nature of the input data has minimal impact on the final outcomes. Notably, testing accuracy reaches 56% when using 10–16 strong entangling layers, highlighting the critical role of entanglement in model performance. For SOTA comparison, we replicate existing quantum self-attention methods and demonstrate that our model consistently achieves 10-23% better performance results. Additionally, we compare different optimizers and find that Nesterov Momentum delivers the most effective convergence for our architecture.

IEEE Globecom Workshops (GC Wkshps)•Dec 8, 2025•0 citations

MedQViT: A Lightweight Hybrid Quantum Vision Transformer for Medical Image Analysis

Amandeep Singh Bhatia, Mandeep Kaur Saggi

Classical Vision Transformers (CViTs) have demonstrated strong performance in medical image analysis, supporting a wide range of digital health applications. However, their self-attention mechanisms can require substantial computational resources and model parameters. In parallel, quantum machine learning (QML) has attracted increasing interest in healthcare and other scientific domains, particularly because quantum-enhanced models may provide expressive representations for small datasets. In this paper, we propose MedQViT, a robust and parameter-efficient Quantum Vision Transformer for medical image classification. To reduce the computational cost of global self-attention, MedQViT combines a shared four-qubit parameterized quantum circuit with a quantum-feature state memory and trainable landmark tokens. This hybrid architecture encodes image patches into compact quantum-enhanced representations while using classical layers for feature aggregation and classification. Experiments on established medical image datasets evaluate the effectiveness and feasibility of MedQViT against classical and quantum baselines. MedQViT achieves competitive performance with substantially fewer parameters than conventional vision models and outperforms several classical and quantum methods in terms of test accuracy (ACC) and area under the receiver operating characteristic curve (AUC).

Proceedings of the AAAI Symposium Series•Nov 6, 2026•0 citations

QiT: Quantum-Inspired Transformer for Visual Recognition Task

Badri N. Patro, Vijay Agneeswaran

Quantum machine learning offers a compelling representational perspective: angle-encoded states inhabit Hilbert spaces in which periodic similarities and interactions can be expressed naturally. Realizing this perspective for visual recognition remains difficult, however, because present quantum neural networks are constrained by limited qubit counts, costly circuit simulation and measurement, noise, and unstable optimization on noisy intermediate-scale quantum devices. We investigate whether useful structural ideas from quantum models can instead be realized as scalable classical Transformer operations. We introduce QiT, a Quantum-inspired Transformer for vision tasks with three components: (i) angle-inspired encoding that maps image tokens to learned trigonometric Hilbert-space features analogous to quantum rotation-based state encoding; (ii) self-attention over these periodic features, inducing a classical cosine kernel approximated to quantum fidelity kernels; and (iii) gated multiplicative emulation, a trainable classical surrogate for interaction terms found in variational circuits. All components are differentiable tensor operations, so QiT claims neither quantum computation nor quantum speedup and retains the $\mathcal{O}(N^2D)$ attention complexity of a standard Vision Transformer. Across image-classification benchmarks, QiT is competitive with a matched classical Transformer while avoiding the severe runtime cost observed for a small simulated quantum Transformer. QiT-B reaches 78.3\% ImageNet-1K top-1 accuracy with 45.7M parameters and 11.5 GFLOPs. These results position QiT as a scalable baseline for isolating and evaluating quantum-motivated inductive biases in visual recognition.

quant-pharXiv:2609.17789•Sep 17, 2026•0 citations

QSAN: A Near-Term Achievable Quantum Self-Attention Network

Jinjing Shi, Ren-Xin Zhao, Wenxuan Wang +2 more

Self-attention mechanism (SAM) is good at capturing the intrinsic connection between features to dramatically boost the performance of machine learning models. Nevertheless, the capability of SAM is not equipped with many current quantum machine learning (QML) models, thus confining their expansion on massive high-dimensional quantum data. To address the above problems, a quantum SAM (QSAM) consisting of a quantum logic similarity (QLS)-based quantum bit self-attention score matrix (QBSASM) is introduced to augment the data representation of SAM exponentially. According to QSAM, the framework and quantum circuits of a one-step achievable quantum self-attention network (QSAN) are designed to consider measurement times compression fully. Moreover, a prototype of quantum coordinates is presented during the design process to describe the mathematical relationship between the output bits and the control bits to facilitate the programming. Ultimately, MNIST binary classification experiments on the PennyLane platform and comparisons with cutting-edge QML models demonstrate QSAN converges about $1.7\times $ and $2.3\times $ faster than hardware-efficient ansatz and quantum approximate optimization algorithm (QAOA) ansatz, respectively, with similar parameter configurations and 100% prediction accuracy, which indicates that it has a better learning capability. In the CIFAR-10 classification experiments, QSAN achieves high prediction accuracy at a small scale relative to classical machine learning models. Predictably, QSAN elevates the efficiency of QML models and lays the foundation for future quantum computers to perform machine learning on massive amounts of data while promoting the advancement of quantum computer vision and other fields.

arXiv:2207.07563•IEEE Transactions on Neural Networks and Learning Systems•Dec 16, 2024•53 citations

Hybrid Neural Network Based on Parameterized Quantum Circuits for Image Classification

Jun Zhang, Xiling Xue, Yue Ruan

Image classification is a core task in the field of computer vision, with significant implications for applications such as medical imaging and autonomous driving. Classical neural network architectures face numerous challenges when dealing with high-dimensional and complex data, particularly in terms of feature extraction and data processing capabilities, limiting their performance in certain applications. Quantum computing offers promising new approaches to processing high-dimensional complex data, most notably through Parameterized Quantum Circuits (PQC) as quantum hybrid neural networks for image classification. In this paper, we propose a hybrid neural network featuring a specific PQC structure designed to generate entanglement in accordance with sample images. This PQC structure utilizes the adjacency matrix of the corresponding graph of the sample image, embedding adjacency relationships among nodes into the entanglement module, thereby generating entanglement aligned with the features of specific images. Numerical results validate the effectiveness of our PQC model across various image classification tasks, enhancing the efficiency of feature learning and representation. Our model demonstrates significant advantages in classification accuracy and learning efficiency on classical datasets. Our research not only advances the theoretical understanding of hybrid quantum networks but also paves the way for practical implementations in real-world scenarios.

Cross Strait Radio Science and Wireless Technology Conference (CSRSWTC)•Nov 4, 2024•3 citations
Transformer now underpins modern AI as its core infrastructure. Its defining capability-dynamically focusing on the most relevant information in complex inputs-is bounded above by the self-attention scoring function. Quantum computing, with its superposition, entanglement, and probabilistic outputs, offers a fundamentally distinct computational framework for exploring beyond the design constraints of classical scoring functions. While quantum attention mechanisms have shown initial promise, existing works remain largely confined to redefining feature similarity measures, leaving the systematic use of parameterized quantum circuits (PQCs) as scoring functions largely unexplored; a substantial portion of existing schemes further rely on purely quantum architectures, precluding effective encoding of high-dimensional image inputs in the Noisy Intermediate-Scale Quantum era. We propose the Quantum Parameterized Self-Attention Network (QPSAN), implementing the self-attention scoring function via PQCs with only 5 trainable quantum parameters per layer. QPSAN computes query-key attention scores through quantum state encoding and joint measurement, yielding naturally bounded outputs without the explicit scaling of classical dot-product attention. We further establish a theoretical framework of the mathematical properties of this scoring function, demonstrating its potential to capture complex nonlinear query-key interactions, and quantifying the structural constraints of the encoding layer via effective degrees of freedom analysis. Experiments on four vision datasets show that QPSAN significantly outperforms the Vision Transformer (ViT) baseline, with the quantum representational advantage amplifying as data complexity increases. Ablation studies indicate that the performance gains may stem from the structural inductive bias of the quantum circuit rather than from parameter scale.
This article proposes a quantum self-attention neural network (QSAN) model that is implementable on parameterized quantum circuits (PQCs), providing a novel avenue for addressing natural language processing (NLP) tasks via quantum computing. The QSAN architecture comprises four blocks: the data preprocessing block, the quantum encoding block, the model design block, and the network optimization block. Through these blocks, classical text data are initially preprocessed and encoded into quantum states. The rich semantic features and intricate relationships within these quantum states are then captured and learned by a quantum self-attention layer and a quantum fully connected layer. Finally, the model performance is enhanced by optimizing the network parameters. Simulation results on publicly available topic classification and sentiment analysis datasets demonstrate that the QSAN model outperforms state-of-the-art baselines. As a limitation of the current study, these evaluations are restricted to a maximum sentence length of eight words due to simulation and hardware constraints. Moreover, the effectiveness, scalability, trainability, and robustness of the proposed model are comprehensively evaluated, highlighting its superior performance and potential for advanced NLP tasks.
Quantum self-attention is a promising approach to replace traditional self-attention mechanisms in classical deep learning, which often require a massive number of GPUs for training. Although several early studies have proposed quantum self-attention architectures, these models often struggle with classification tasks in vision transformers. In this study, we address this limitation by enhancing quantum self-attention through integration with circuit classifier design and, crucially, by optimizing the number of strong entangling layers. Evaluation results show that incorporating an optimal number of entangling layers substantially improves classification performance, outperforming models that rely solely on directly measured quantum self-attention. We further analyze how varying the number of quantum classifier layers affects classification accuracy during training. Experiments on downscaled MNIST and MC datasets reveal that the nature of the input data has minimal impact on the final outcomes. Notably, testing accuracy reaches 56% when using 10–16 strong entangling layers, highlighting the critical role of entanglement in model performance. For SOTA comparison, we replicate existing quantum self-attention methods and demonstrate that our model consistently achieves 10-23% better performance results. Additionally, we compare different optimizers and find that Nesterov Momentum delivers the most effective convergence for our architecture.
Classical Vision Transformers (CViTs) have demonstrated strong performance in medical image analysis, supporting a wide range of digital health applications. However, their self-attention mechanisms can require substantial computational resources and model parameters. In parallel, quantum machine learning (QML) has attracted increasing interest in healthcare and other scientific domains, particularly because quantum-enhanced models may provide expressive representations for small datasets. In this paper, we propose MedQViT, a robust and parameter-efficient Quantum Vision Transformer for medical image classification. To reduce the computational cost of global self-attention, MedQViT combines a shared four-qubit parameterized quantum circuit with a quantum-feature state memory and trainable landmark tokens. This hybrid architecture encodes image patches into compact quantum-enhanced representations while using classical layers for feature aggregation and classification. Experiments on established medical image datasets evaluate the effectiveness and feasibility of MedQViT against classical and quantum baselines. MedQViT achieves competitive performance with substantially fewer parameters than conventional vision models and outperforms several classical and quantum methods in terms of test accuracy (ACC) and area under the receiver operating characteristic curve (AUC).
Quantum machine learning offers a compelling representational perspective: angle-encoded states inhabit Hilbert spaces in which periodic similarities and interactions can be expressed naturally. Realizing this perspective for visual recognition remains difficult, however, because present quantum neural networks are constrained by limited qubit counts, costly circuit simulation and measurement, noise, and unstable optimization on noisy intermediate-scale quantum devices. We investigate whether useful structural ideas from quantum models can instead be realized as scalable classical Transformer operations. We introduce QiT, a Quantum-inspired Transformer for vision tasks with three components: (i) angle-inspired encoding that maps image tokens to learned trigonometric Hilbert-space features analogous to quantum rotation-based state encoding; (ii) self-attention over these periodic features, inducing a classical cosine kernel approximated to quantum fidelity kernels; and (iii) gated multiplicative emulation, a trainable classical surrogate for interaction terms found in variational circuits. All components are differentiable tensor operations, so QiT claims neither quantum computation nor quantum speedup and retains the O(N2D)\mathcal{O}(N^2D)O(N2D) attention complexity of a standard Vision Transformer. Across image-classification benchmarks, QiT is competitive with a matched classical Transformer while avoiding the severe runtime cost observed for a small simulated quantum Transformer. QiT-B reaches 78.3\% ImageNet-1K top-1 accuracy with 45.7M parameters and 11.5 GFLOPs. These results position QiT as a scalable baseline for isolating and evaluating quantum-motivated inductive biases in visual recognition.
Self-attention mechanism (SAM) is good at capturing the intrinsic connection between features to dramatically boost the performance of machine learning models. Nevertheless, the capability of SAM is not equipped with many current quantum machine learning (QML) models, thus confining their expansion on massive high-dimensional quantum data. To address the above problems, a quantum SAM (QSAM) consisting of a quantum logic similarity (QLS)-based quantum bit self-attention score matrix (QBSASM) is introduced to augment the data representation of SAM exponentially. According to QSAM, the framework and quantum circuits of a one-step achievable quantum self-attention network (QSAN) are designed to consider measurement times compression fully. Moreover, a prototype of quantum coordinates is presented during the design process to describe the mathematical relationship between the output bits and the control bits to facilitate the programming. Ultimately, MNIST binary classification experiments on the PennyLane platform and comparisons with cutting-edge QML models demonstrate QSAN converges about $1.7\times $ and $2.3\times $ faster than hardware-efficient ansatz and quantum approximate optimization algorithm (QAOA) ansatz, respectively, with similar parameter configurations and 100% prediction accuracy, which indicates that it has a better learning capability. In the CIFAR-10 classification experiments, QSAN achieves high prediction accuracy at a small scale relative to classical machine learning models. Predictably, QSAN elevates the efficiency of QML models and lays the foundation for future quantum computers to perform machine learning on massive amounts of data while promoting the advancement of quantum computer vision and other fields.
Image classification is a core task in the field of computer vision, with significant implications for applications such as medical imaging and autonomous driving. Classical neural network architectures face numerous challenges when dealing with high-dimensional and complex data, particularly in terms of feature extraction and data processing capabilities, limiting their performance in certain applications. Quantum computing offers promising new approaches to processing high-dimensional complex data, most notably through Parameterized Quantum Circuits (PQC) as quantum hybrid neural networks for image classification. In this paper, we propose a hybrid neural network featuring a specific PQC structure designed to generate entanglement in accordance with sample images. This PQC structure utilizes the adjacency matrix of the corresponding graph of the sample image, embedding adjacency relationships among nodes into the entanglement module, thereby generating entanglement aligned with the features of specific images. Numerical results validate the effectiveness of our PQC model across various image classification tasks, enhancing the efficiency of feature learning and representation. Our model demonstrates significant advantages in classification accuracy and learning efficiency on classical datasets. Our research not only advances the theoretical understanding of hybrid quantum networks but also paves the way for practical implementations in real-world scenarios.