ABSTRACT We present a hybrid quantum–classical convolutional neural network (QC‐CNN) for classifying orbital angular momentum (OAM) modes in free‐space optical (FSO) channels affected by atmospheric turbulence. The approach embeds a convolutional quantum (quanvolutional) preprocessing layer that encodes 2 × 2 image patches into parameterized quantum circuits with rotation and entangling operations, yielding quantum‐extracted features that feed into conventional convolutional layers. We employ a two‐stage training protocol: large‐scale pretraining on 940 000 intensity images spanning a range of turbulence strengths, followed by task‐specific fine‐tuning with selective layer freezing. Turbulence was modeled using a modified Hill–Andrews phase screen and evaluated at multiple strengths over 1000 m propagation distance. The hybrid model achieves high classification accuracies (>97% across strong turbulence scenarios) and reduces bit‐error rates (BER) from up to 0.11 without pretraining to 0.001–0.08 after quantum‐layer pretraining. Results demonstrate that quantum‐inspired preprocessing combined with transfer learning significantly improves resilience to severe atmospheric distortions and offers a scalable route toward real‐time, high‐fidelity OAM demodulation in FSO systems.
ABSTRACT We present a hybrid quantum–classical convolutional neural network (QC‐CNN) for classifying orbital angular momentum (OAM) modes in free‐space optical (FSO) channels affected by atmospheric turbulence. The approach embeds a convolutional quantum (quanvolutional) preprocessing layer that encodes 2 × 2 image patches into parameterized quantum circuits with rotation and entangling operations, yielding quantum‐extracted features that feed into conventional convolutional layers. We employ a two‐stage training protocol: large‐scale pretraining on 940 000 intensity images spanning a range of turbulence strengths, followed by task‐specific fine‐tuning with selective layer freezing. Turbulence was modeled using a modified Hill–Andrews phase screen and evaluated at multiple strengths over 1000 m propagation distance. The hybrid model achieves high classification accuracies (>97% across strong turbulence scenarios) and reduces bit‐error rates (BER) from up to 0.11 without pretraining to 0.001–0.08 after quantum‐layer pretraining. Results demonstrate that quantum‐inspired preprocessing combined with transfer learning significantly improves resilience to severe atmospheric distortions and offers a scalable route toward real‐time, high‐fidelity OAM demodulation in FSO systems.
Abstract Quantum machine learning (QML) offers a breakthrough by producing atypical patterns that classical systems cannot efficiently produce. This study explores quantum neural networks (QNN) for classifying orbital angular momentum (OAM) modes. A fully quantum model achieving 99.8% accuracy across 10 OAM classes and a hybrid ConvQuantum model reaching accuracy levels of 99.8% under atmospheric turbulence strength (AT) of is developed. The ConvQuantum model significantly reduces prediction time, averaging 0.00025 s per sample, which is essential for real‐time free space optical (FSO) communication. To demonstrate quantum technology's potential in OAM‐FSO communication, a 100 × 100 grayscale image is successfully transmitted, and compared the results with several known models in the field, including feedforward neural networks (FNN), convolutional neural network (CNN), and a vision transformer (ViT) model. The findings indicate that the QNN models outperform classical approaches in calculation time, showcasing the efficiency and enhanced processing power of quantum computing in OAM‐FSO communication.
Quantum computing is emerging as a transformative tool for communication systems, offering the potential to overcome long-standing physical limitations. In free-space optical networks, orbital angular momentum (OAM) multiplexing promises massive capacity gains, but its practical use is fundamentally constrained by multiphysics degradations such as atmospheric turbulence, volumetric Mie scattering, and stochastic quantum noise. These effects induce nonlinear modal crosstalk and severe beam distortions, against which classical approaches—most notably convolutional neural networks (CNNs)—provide only partial and non-scalable compensation. To address this gap, we report the first use of variational quantum neural networks (QNNs) for adaptive OAM beam compensation in realistic channels. By embedding parameterized entangling layers into a supervised regression pipeline, our QNN achieves end-to-end reconstruction of distorted Laguerre–Gaussian beams with topological charges l ∈ {1,4,8,12}. Using experimentally validated channel parameters, QNNs achieve mean squared error as low as 4.0 × 10− 6, SSIM above 0.99, and bit-error rates suppressed by > 99.9% (0.0125% BER). To ensure scalability, we introduce the quasi-quantum neural network (QqNN), a classical surrogate that emulates quantum dynamics via tensorial projections, achieving near-optimal performance (0.0375% BER) at reduced complexity. This hybrid framework positions QNNs as a quantum-resilient paradigm for OAM decoding and establishes QqNNs as the first scalable surrogate for near-term deployment.
This project presents a comparative study between a classical convolutional neural network (CNN) and a hybrid quantum neural network (QNN) for orbital angular momentum (OAM) image classification. The analysis focuses on parameter efficiency, showing that the QNN achieves competitive (and in some cases improved) classification performance while using significantly fewer trainable parameters—approximately 8× fewer than the CNN (1,204 vs. 9,928 parameters). This reduction arises from fundamental differences in representation: CNNs rely on large numbers of localized weights in convolutional and fully connected layers, whereas QNNs use parameterized quantum circuits in which a small set of rotation angles governs the evolution of quantum states in a high-dimensional Hilbert space. Through quantum properties such as superposition, entanglement, and global unitary transformations, QNNs enable compact yet expressive feature representations, effectively capturing complex correlations with fewer parameters. The comparison is conducted under controlled conditions, including identical dataset, preprocessing, training splits, and optimization settings, ensuring fairness. The results highlight that model capacity is not solely determined by parameter count, and demonstrate the potential of hybrid quantum models for parameter-efficient learning in challenging visual classification tasks involving distorted OAM modes. cite following paper if you are using this codes: Manavalan, G., Arnon, S. Quantum neural network-based compensation of distorted orbital angular momentum beams in complex media. Sci Rep 16, 1361 (2026). https://doi.org/10.1038/s41598-025-31021-1 Manavalan, G., Arnon, S. Decrypting chaotic visual ciphers via quasi quantum neural networks (Q²NNs). Sci Rep 16, 9937 (2026). https://doi.org/10.1038/s41598-026-41513-3
A high-dimensional quantum key distribution (QKD), which adopts degrees of freedom of the orbital angular momentum (OAM) states, is beneficial to realize secure and high-speed QKD. However, the helical phase of a vortex beam that carries OAM is sensitive to the atmospheric turbulence and easily distorted. In this paper, an adaptive compensation method using deep learning technology is developed to improve the performance of OAM-encoded QKD schemes. A convolutional neural network model is first trained to learn the mapping relationship of intensity profiles of inputs and the turbulent phase, and such mapping is used as feedback to control a spatial light modulator to generate a phase screen to correct the distorted vortex beam. Then an OAM-encoded QKD scheme with the capability of real-time phase correction is designed, in which the compensation module only needs to extract the intensity distributions of the Gaussian probe beam and thus ensures that the information encoded on OAM states would not be eavesdropped. The results show that our method can efficiently improve the mode purity of the encoded OAM states and extend the secure distance for the involved QKD protocols in the free-space channel, which is not limited to any specific QKD protocol.
Image classification plays a critical role across various industries, yet the rapid growth of visual data has significantly increased computational complexity. Convolutional neural networks (CNNs) have demonstrated strong capabilities in efficient feature extraction, but they still face trade‐offs between performance and computational cost when handling large‐scale data. Recent studies have explored integrating quantum mechanical principles into classical CNNs to enhance model expressivity. However, existing hybrid quantum–classical convolutional architectures remain limited in terms of feature capture and computational efficiency. In this study, we propose a novel hybrid quantum–classical convolutional neural network (QC‐CNN‐Parallel), which introduces parallel quantum and classical convolutional branches and extends the concept of multiscale convolution to hybrid networks. This design enables the capture of larger contextual information during convolution while reducing circuit depth and computational overhead. Moreover, we employ parameterized quantum circuits (PQCs) with expressibility, entanglement, and discreteness metrics to select optimal quantum circuits and further enhance model performance. We evaluate QC‐CNN‐Parallel on three grayscale image datasets: MNIST, Fashion‐MNIST, and Overhead‐MNIST. Experimental results show that our model improves average training accuracy by 4.89%, 2.77%, and 6.24% over existing hybrid quantum CNNs and classical CNNs, respectively, while exhibiting superior robustness. These results demonstrate that QC‐CNN‐Parallel achieves enhanced classification performance with high computational efficiency and practical potential.
Quantum machine learning has emerged as a promising approach for enhancing classical deep learning models by leveraging quantum computational properties.In this work, a hybrid Quantum-Classical Convolutional Neural Network (QC-CNN) is proposed for overhead image classification using the Overhead-MNIST dataset.The model integrates a classical Convolutional Neural Network (CNN) for spatial feature extraction with a parameterized quantum circuit that performs quantum feature transformation through angle embedding and strongly entangling layers.Initially, input images are preprocessed and resized, after which the CNN extracts high-level feature representations.These features are then encoded into quantum states and processed through a variational quantum layer to capture complex feature correlations using quantum operations.The output of the quantum circuit is passed to fully connected layers for final classification.The hybrid architecture is trained end-to-end using a gradient-based optimization approach.Experimental results demonstrate that the proposed QC-CNN model achieves high classification accuracy on the selected classes of the Overhead-MNIST dataset while maintaining stable training behavior.The results indicate that integrating quantum circuits with classical deep learning frameworks can improve feature representation and provide a potential pathway toward future quantum-enhanced intelligent vision systems.