计算机科学
生成语法
对抗制
人工智能
生成对抗网络
机器学习
生成模型
深度学习
出处
期刊:International Conference on Cloud Computing
日期:2020-01-01
被引量:1
标识
DOI:10.1109/confluence47617.2020.9058040
摘要
From the past few years, Generative adversarial networks (GANs) have gained more and more interest of researchers of Artificial Intelligence and this is only due to the reliability on huge amount of data, well designed network architectures and smart training techniques because of which they produce highly realistic pieces of content of images, texts and sounds. The inspirational idea of working in GANs has been derived from game theory, named as the zero–sum game. GANs consist of two components-a generator as well as a discriminator both of which act like two players of the game playing in opposition with each other. This paper focuses on the basic theory and principle mechanism of GANs. Next, the paper discusses few variants based on architecture as well as loss functions of some kinds. Finally, the last section of paper presents few other variants of GANs which are implemented in the field of computer vision and other real world problems. It is found that this area has a wider scope in terms of virtual real interaction and integration along with parallel learning. So it is considered as new implementation area for GANs in the coming future.
科研通智能强力驱动
Strongly Powered by AbleSci AI