生成对抗网络模型的基本介绍和应用综述

A Basic Introduction and Application Review of Generative Adversarial Network's Model

生成对抗网络(GAN)目前已然成为人工智能领域的一个研究热点,它为无监督学习的研究带来新的活力,对生成式模型的发展具有举足轻重的意义。针对生成对抗网络这一热点模型,介绍生成对抗网络基本原理,以及在其基础上改进的几种典型的变体:条件生成对抗网络(CGAN)、深度卷积对抗网络(DCGAN)、循环一致生成网络(CycleGAN)以及堆积生成对抗网络(StackGAN)。同时还介绍生成对抗网络在计算机视觉领域中的几种应用,对生成对抗网络做总结和展望。

The generation of Generative Adversarial Networks(GAN)has become a research hotspot in the field of artificial intelligence,it brings new vitality to the study of unsupervised learning,and plays a pivotal role in the development of the generative model.In order to generate a hotspot model against the network,mainly introduces several typical variants of generating the basic principles of the anti-network and improving it on the basis of the Conditional Generation Confrontation Network(CGAN)and the Deep Convolutional Confrontation Network(DCGAN).Cycle Consistent Generation Network(CycleGAN)and Stack Generation Confrontation Network(StackGAN).At the same time,introduces several applications of generating confrontation networks in the field of computer vision.Finally,summarizes and forecasts the generation of confrontation networks.