Scholarlink

<p>Quantum generative adversarial networks (QGANs), as a cutting-edge direction in quantum machine learning, show potential for quantum data modeling and generation tasks, though practical advantages over classical methods remain to be fully established. By integrating the adversarial training mechanism with the characteristics of quantum computing, QGANs provide new perspectives for representing and learning complex distributions. This paper presents a systematic review of the development of QGANs. It first revisits the fundamental principles of generative adversarial networks and variational quantum algorithms, then introduces the core structure and training process of QGANs, analyzes the role of different loss functions in model optimization, and summarizes the design ideas behind various QGANs architectures. Building on this foundation, the paper compares experimental results of existing QGANs models, outlines the commonly used datasets and evaluation metrics, and offers a comprehensive comparison of their performance across applications in finance, healthcare, and image generation. Furthermore, the paper discusses the potential prospects of QGANs in practical scenarios based on the current research landscape, and suggests future directions in establishing standardized benchmarking systems, developing resource-efficient training methods, and identifying realistic application domains, aiming to provide guidance and inspiration for the future development of QGANs.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Quantum generative adversarial networks: a comprehensive survey of theories, applications, and challenges in the NISQ era

  • 期刊论文

  • 发表日期：

    2026年5月22日

查看全文
  • Han Qi,
  • Yihan Xu,
  • Hao Wang,
  • Abdullah Gani,
  • Lip Yee Por

摘要

Quantum generative adversarial networks (QGANs), as a cutting-edge direction in quantum machine learning, show potential for quantum data modeling and generation tasks, though practical advantages over classical methods remain to be fully established. By integrating the adversarial training mechanism with the characteristics of quantum computing, QGANs provide new perspectives for representing and learning complex distributions. This paper presents a systematic review of the development of QGANs. It first revisits the fundamental principles of generative adversarial networks and variational quantum algorithms, then introduces the core structure and training process of QGANs, analyzes the role of different loss functions in model optimization, and summarizes the design ideas behind various QGANs architectures. Building on this foundation, the paper compares experimental results of existing QGANs models, outlines the commonly used datasets and evaluation metrics, and offers a comprehensive comparison of their performance across applications in finance, healthcare, and image generation. Furthermore, the paper discusses the potential prospects of QGANs in practical scenarios based on the current research landscape, and suggests future directions in establishing standardized benchmarking systems, developing resource-efficient training methods, and identifying realistic application domains, aiming to provide guidance and inspiration for the future development of QGANs.