一种学习多域数据联合分布的量子耦合生成对抗网络

打开文本图片集

中图分类号： TN929.5 文献标志码：A DOI： 10.12305/j.issn.1001-506X.2026.03.05

Abstract： Ordinary quantum generative adversarial network（GAN） can only handle single domain data and are limited by the number ofquantum bits，resulting in a significant decrease in generation quality.Inregard to this，aquantumcoupled GAN（QCoGAN） is proposed to learn the joint distribution of multi domain data. QCoGAN combines the parallel computing capability ofquantumcomputing with the leaing abilityof classical GANs tocoupleand optimize thestructureofquantum GANs. Compared with ordinary GANs，itcan capture more image details.By applying weight sharing constraints between quantum parameter layers，it efectivelycontrols the network capacityand improves training eficiency.Applying quantum GAN to domain adaptation problems and QCoGAN to joint distributed learming tasks on multiple handwriten datasets is demonstrated its superiority in handling multi domain data and has broad application prospects.

Keywords：quantum computing；machine learming；generative adversarial network （GAN）；quantum coupling; multi-domain data

0 引言

机器学习和量子物理学之间的相互作用可能会为这两个领域带来前所未有的前景1]。（剩余12771字）