《计算机应用研究》|Application Research of Computers

生成对抗网络GAN的研究进展

Research progress on generative adversarial network

免费全文下载 (已被下载 次)  
获取PDF全文
作者 张恩琪,顾广华,赵晨,赵志明
机构 1.燕山大学 信息科学与工程学院,河北 秦皇岛 066004;2.河北省信息传输与信号处理重点实验室,河北 秦皇岛 066004
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2021)04-002-0968-07
DOI 10.19734/j.issn.1001-3695.2020.05.0095
摘要 基于零和博弈思想的生成式对抗网络(generative adversarial network,GAN)模型的意义在于可通过无监督学习获得数据的分布,并能生成较逼真的数据。它可以应用在很多领域,尤其是在计算机视觉领域中的图像生成方面取得了很大成果,成为当下的一个研究热点。以GAN模型及其在特定领域的应用结果为研究对象,对GAN的改进和扩展的研究成果进行了广泛的研究,并从图像超分辨率重建、文本合成图片等多个实际应用领域展开讨论,系统地梳理、总结出GAN的优势与不足,同时结合自然语言处理、强化学习对GAN的发展趋势及应用前景进行预测分析。
关键词 零和博弈思想; 生成式对抗网络; 无监督学习; 图像超分辨率重建; 文本合成图片
基金项目 河北省自然科学基金资助项目(F2017203169)
河北省高等学校科学研究重点项目(ZD2017080)
本文URL http://www.arocmag.com/article/01-2021-04-002.html
英文标题 Research progress on generative adversarial network
作者英文名 Zhang Enqi, Gu Guanghua, Zhao Chen, Zhao Zhiming
机构英文名 1.School of Information Science & Engineering,Yanshan University,Qinhuangdao Hebei 066004,China;2.Key Laboratory of Information Transmission & Signal Processing of Hebei Province,Qinhuangdao Hebei 066004,China
英文摘要 The significance of the GAN model based on the zero-sum game idea is that the data distribution can be obtained through unsupervised learning, and it can generate more realistic data. It can be applied in many fields, especially the field of computer vision, and achieved great results in image generation, so that it becomes a hot spot in current research. This paper took the GAN model and its application results in specific fields as the research object, conducted extensive research on the improvement and expansion of GAN research results, and discussed from multiple practical application areas such as image super-resolution reconstruction and text synthesis pictures. It systematically sorted out and summarized the advantages and disadvantages of GAN, predicted and analyzed the development trend and application prospect of GAN in combination with natural language processing and reinforcement learning.
英文关键词 zero-sum game theory; generative adversarial network; unsupervised learning; image super-resolution reconstruction; text synthesis picture
参考文献 查看稿件参考文献
 
收稿日期 2020/5/11
修回日期 2020/7/2
页码 968-974
中图分类号 TG333
文献标志码 A