We propose a novel approach to generative adversarial networks (GANs) in which the standard i.i.d. Gaussian latent prior is replaced or hybridized with a quantum-correlated prior derived from measurements of a 16-qubit entangling circuit. Each latent sample is generated by grouping repeated shots per qubit into a binary fraction, applying the inverse Gaussian CDF to obtain a 16-dimensional Gaussian vector whose joint copula reflects genuine quantum entanglement, and then projecting into the high-dimensional space via a fixed random matrix. By pre-sampling tens of millions of bitstrings, either from a noiseless simulator or from IBM hardware, we build large pools of independent but internally quantum-correlated latents. We integrate this prior into three representative architectures (WGAN, SNGAN, BigGAN) on CIFAR-10, making no changes to the neural network structure or training hyperparameters. The hybrid latent representations incorporating hardware-derived noise consistently lower the FID relative to both the classical baseline and the simulator variant, especially when the quantum component constitutes a substantial fraction of the prior. In addition, we execute on the QPU in parallel to not only save computing time but also further decrease the FID up to 17% in BigGAN. These results indicate that intrinsic quantum randomness and device-specific imperfections can provide a structured inductive bias that enhances GAN performance. Our work demonstrates a practical pipeline for leveraging noisy quantum hardware to enrich deep-generative modeling, opening a new interface between quantum information and machine learning. All code and data are available at https://github.com/Neon8988/GAN_QN.git.
We propose a novel approach to generative adversarial networks (GANs) in which the standard i.i.d. Gaussian latent prior is replaced or hybridized with a quantum-correlated prior derived from measurements of a 16-qubit entangling circuit. Each latent sample is generated by grouping repeated shots per qubit into a binary fraction, applying the inverse Gaussian CDF to obtain a 16-dimensional Gaussian vector whose joint copula reflects genuine quantum entanglement, and then projecting into the high-dimensional space via a fixed random matrix. By pre-sampling tens of millions of bitstrings, either from a noiseless simulator or from IBM hardware, we build large pools of independent but internally quantum-correlated latents. We integrate this prior into three representative architectures (WGAN, SNGAN, BigGAN) on CIFAR-10, making no changes to the neural network structure or training hyperparameters. The hybrid latent representations incorporating hardware-derived noise consistently lower the FID relative to both the classical baseline and the simulator variant, especially when the quantum component constitutes a substantial fraction of the prior. In addition, we execute on the QPU in parallel to not only save computing time but also further decrease the FID up to 17% in BigGAN. These results indicate that intrinsic quantum randomness and device-specific imperfections can provide a structured inductive bias that enhances GAN performance. Our work demonstrates a practical pipeline for leveraging noisy quantum hardware to enrich deep-generative modeling, opening a new interface between quantum information and machine learning. All code and data are available at https://github.com/Neon8988/GAN_QN.git.
Abstract Quantum machine learning offers powerful advantages in processing high-dimensional data, but loading classical probability distributions into quantum states remains a major challenge. This work extends quantum generative adversarial networks (qGANs) to overcome two critical limitations of early approaches: poor robustness to noise on near-term devices and limited capacity to model complex, multi-modal distributions. We introduce hybrid qGAN architectures that combine Wasserstein GAN with gradient penalty (WGAN-GP) and maximum mean discrepancy (MMD) losses with expressive quantum circuits, including quantum convolutional neural networks (QCNNs) and EfficientSU2 ansätze. Training is performed using seamless PyTorch–Qiskit integration, enabling stable optimization on both simulators and real quantum hardware.Experiments on 2D Gaussian and log-normal distributions—directly relevant to financial modeling—show faster convergence and higher fidelity than prior qGAN designs, with up to 80% lower Wasserstein distance under 5% depolarizing noise. The framework is further validated through European call option pricing via quantum amplitude estimation (QAE), achieving pricing errors below 1% on IBM’s 20-qubit superconducting systems. Theoretical analysis confirms gradient stability under noise, while empirical results include comprehensive noise sweeps, t-SNE latent space visualizations, and rigorous statistical tests (KL divergence, KS tests). This work establishes a practical path toward quantum advantage in generative modeling, with fully open-source code for reproducibility.
Generative adversarial network (GAN) is one of the widely-adopted machine-learning frameworks for a wide range of applications such as generating high-quality images, video, and audio contents. However, training a GAN could become computationally expensive for large neural networks. In this work, we propose a hybrid quantum-classical architecture for improving GAN (denoted as QC-GAN). The performance was examed numerically by benchmarking with a classical GAN using MindSpore Quantum on the task of hand-written image generation. The generator of the QC-GAN consists of a quantum variational circuit together with a one-layer neural network, and the discriminator consists of a traditional neural network. Leveraging the entangling and expressive power of quantum circuits, our hybrid architecture achieved better performance (Frechet Inception Distance) than the classical GAN, with much fewer training parameters and number of iterations for convergence. We have also demonstrated the superiority of QC-GAN over an alternative quantum GAN, namely pathGAN, which could hardly generate 16$\times$16 or larger images. This work demonstrates the value of combining ideas from quantum computing with machine learning for both areas of Quantum-for-AI and AI-for-Quantum.
Generative Adversarial Networks (GANs) have demonstrated immense potential in synthesizing diverse and high-fidelity images; however, critical questions remain regarding how quantum principles might best enhance their representational capacity. This thesis investigates hybrid quantum-classical GAN architectures supplemented by transfer learning in the discriminator, systematically examining whether incorporating Variational Quantum Circuits (VQCs) into the generator, the discriminator, or both improves performance over fully classical baselines. We evaluate each configuration using standard GAN benchmarks (CIFAR-10 classes: Birds, Cars, and Dogs) under varying sample sizes (5,000 vs. 2,500 images per class) to gauge both robustness and data efficiency. Our findings indicate that fully hybrid models, which incorporate VQCs in both the generator and the discriminator, consistently produce images of higher visual quality and achieve more favorable quantitative metrics (such as Fréchet Inception Distance (FID), Kernel Inception Distance (KID), and Inception Score (IS)) compared to their fully classical counterpart. In particular, VQCs in the generator accelerate early feature learning, whereas those in the discriminator, despite exhibiting slower initial convergence, ultimately facilitate more refined synthetic outputs. Moreover, the model sustains near-comparable performance even when the dataset size is halved, suggesting that transfer learning and quantum enhancements mitigate data limitations. By comparing classical and hybrid configurations in detail, this thesis illustrates the trade-offs arising from VQCs in GANs, such as variations in early-epoch convergence and final generative quality. Overall, the results underscore that carefully integrating quantum computing principles with classical adversarial training and pretrained feature extraction can significantly enrich GAN-based image synthesis. These insights open avenues for future work on higher-resolution tasks, alternative quantum circuit designs, and experimentation with emerging quantum hardware.
One of the most promising applications in the era of Noisy Intermediate-Scale Quantum (NISQ) computing is quantum generative adversarial networks (QGANs), which offer significant quantum advantages over classical machine learning in various domains. However, QGANs suffer from mode collapse and lack explicit control over the features of generated outputs. To overcome these limitations, we propose InfoQGAN, a novel quantum-classical hybrid generative adversarial network that integrates the principles of InfoGAN with a QGAN architecture. Our approach employs a variational quantum circuit for data generation, a classical discriminator, and a Mutual Information Neural Estimator (MINE) to explicitly optimize the mutual information between latent codes and generated samples. Numerical simulations on synthetic 2D distributions and Iris dataset augmentation demonstrate that InfoQGAN effectively mitigates mode collapse while achieving robust feature disentanglement in the quantum generator. By leveraging these advantages, InfoQGAN not only enhances training stability but also improves data augmentation performance through controlled feature generation. These results highlight the potential of InfoQGAN as a foundational approach for advancing quantum generative modeling in the NISQ era.
Generative Adversarial Networks (GANs) have demonstrated immense potential in synthesizing diverse and high-fidelity images; however, critical questions remain regarding how quantum principles might best enhance their representational capacity. This thesis investigates hybrid quantum-classical GAN architectures supplemented by transfer learning in the discriminator, systematically examining whether incorporating Variational Quantum Circuits (VQCs) into the generator, the discriminator, or both improves performance over fully classical baselines. We evaluate each configuration using standard GAN benchmarks (CIFAR-10 classes: Birds, Cars, and Dogs) under varying sample sizes (5,000 vs. 2,500 images per class) to gauge both robustness and data efficiency. Our findings indicate that fully hybrid models, which incorporate VQCs in both the generator and the discriminator, consistently produce images of higher visual quality and achieve more favorable quantitative metrics (such as Fréchet Inception Distance (FID), Kernel Inception Distance (KID), and Inception Score (IS)) compared to their fully classical counterpart. In particular, VQCs in the generator accelerate early feature learning, whereas those in the discriminator, despite exhibiting slower initial convergence, ultimately facilitate more refined synthetic outputs. Moreover, the model sustains near-comparable performance even when the dataset size is halved, suggesting that transfer learning and quantum enhancements mitigate data limitations.By comparing classical and hybrid configurations in detail, this thesis illustrates the trade-offs arising from VQCs in GANs, such as variations in early-epoch convergence and final generative quality. Overall, the results underscore that carefully integrating quantum computing principles with classical adversarial training and pretrained feature extraction can significantly enrich GAN-based image synthesis. These insights open avenues for future work on higher-resolution tasks, alternative quantum circuit designs, and experimentation with emerging quantum hardware.
Intrusion Detection Systems (IDS) are vital for securing modern networks, but traditional machine learning-based anomaly detection methods often struggle with high-dimensional and imbalanced data. Although classical GANs help improve data generation, they are computationally intensive and limited in scalability. Quantum GANs (QGANs) present a promising alternative, but existing approaches are tested only in simulations. Instead, using real devices presents challenges not only because of the impact of quantum noise but, more significantly, due to usage restrictions imposed by their high cost. To cope with this, we propose a lightweight but effective hybrid QGAN for network intrusion detection, adopting a hybrid training approach by switching between an IBM real quantum device and its noisy simulator version, balancing computational feasibility and relevance in the real world. The results show that our model performs consistently, with minimal performance loss on real hardware.