The paper provides a comprehensive review of Generative Adversarial Networks (GANs), a novel framework for training generative models in an adversarial setup. The authors introduce various deep generative models, the basic theory and training mechanism of GANs, and the latent space. They discuss several representative variants of GANs and highlight the high instability and convergence issues in GAN training. The paper explores the sources of these issues from statistics, game theory, and control theory, and proposes techniques for stable training. It also addresses the challenges in evaluating GANs due to the lack of consensus on suitable evaluation measures and provides a discussion on quantitative and qualitative evaluation metrics. The authors conduct experiments to compare different GAN variants using these metrics and briefly discuss the application areas of GANs. Finally, they outline important open issues and future research trends in GANs.The paper provides a comprehensive review of Generative Adversarial Networks (GANs), a novel framework for training generative models in an adversarial setup. The authors introduce various deep generative models, the basic theory and training mechanism of GANs, and the latent space. They discuss several representative variants of GANs and highlight the high instability and convergence issues in GAN training. The paper explores the sources of these issues from statistics, game theory, and control theory, and proposes techniques for stable training. It also addresses the challenges in evaluating GANs due to the lack of consensus on suitable evaluation measures and provides a discussion on quantitative and qualitative evaluation metrics. The authors conduct experiments to compare different GAN variants using these metrics and briefly discuss the application areas of GANs. Finally, they outline important open issues and future research trends in GANs.