Research On Quantum Generative Algorithm Of Generative Adversarial Networks

Combining quantum resources with classical computing resources to develop more effective algorithms is one of the most promising research directions in computer science.In the quantum-classical hybrid framework,existing quantum devices can be used to the greatest extent for practical applications.In this context,quantum machine learning,as an interdisciplinary subject combining quantum computing and classical machine learning,has attracted more and more attention from governments and researchers.Many scholars have begun to consider combining the advantages of parallel computing capabilities and uncertainty of quantum computing with machine learning tasks to propose new algorithms with special applications.Generative adversarial networks provide a way to learn high-dimensional probability distributions without a large amount of labeled training data.The quantum computing community has recently begun to study different quantum generation algorithms to learn the potential generation methods of high-dimensional quantum states.This paper uses the potential advantages of quantum computing to study the quantum generalization of the generative adversarial network,designs the quantum generative adversarial network algorithm,improves the basic network structure of the generator and the discriminator in the quantum generative adversarial network,and expands the quantum generative adversarial algorithm's research and application in the field of noise and quantum image processing.The research results of this article are summarized as follows:1.Aiming at the difficulty of training the quantum generative adversarial network,a quantum generative adversarial network algorithm based on the quantum Chernoff bound is proposed.The basic idea is to introduce quantum Chernoff bound to improve its loss function under the condition that the structure of the quantum generation adversarial network remains unchanged.Experiments show that the improved quantum generative adversarial network not only improves the convergence speed,but also improves the accuracy of the final result.2.Aiming at the shortcomings of the complex structure of the quantum circuit of the quantum generation network,the circuit composition of the basic building block in the quantum circuit of the quantum generative adversarial network is improved.The quantum gate analysis gradient reduces the number of quantum gates of the basic building block from15 to 13.Through numerical simulation of the influence of quantum noise,the anti noise ability of quantum circuits is studied and analyzed.3.Aiming at the shortcoming that the quantum generative adversarial network can only generate quantum states without combining with classical image data,a quantum generative adversarial network based on quantum image learning and generation is proposed.The basic idea is: by designing the encoding method of the classical image and the quantum state,the classical image is encoded into the quantum state,and then the quantum image is learned and generated by the quantum generation adversarial network.This method can be used not only for the generation of known images,but also for learning unknown quantum images generated by other programs.