{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/gang-of-gans-generative-adversarial-networks","title":"Gang of GANs: Generative Adversarial Networks with Maximum Margin Ranking","arxiv_id":"1704.04865","date":"2017-04-17","proceeding":null,"authors":["Felix Juefei-Xu","Vishnu Naresh Boddeti","Marios Savvides"],"abstract":"Traditional generative adversarial networks (GAN) and many of its variants\nare trained by minimizing the KL or JS-divergence loss that measures how close\nthe generated data distribution is from the true data distribution. A recent\nadvance called the WGAN based on Wasserstein distance can improve on the KL and\nJS-divergence based GANs, and alleviate the gradient vanishing, instability,\nand mode collapse issues that are common in the GAN training. In this work, we\naim at improving on the WGAN by first generalizing its discriminator loss to a\nmargin-based one, which leads to a better discriminator, and in turn a better\ngenerator, and then carrying out a progressive training paradigm involving\nmultiple GANs to contribute to the maximum margin ranking loss so that the GAN\nat later stages will improve upon early stages. We call this method Gang of\nGANs (GoGAN). We have shown theoretically that the proposed GoGAN can reduce\nthe gap between the true data distribution and the generated data distribution\nby at least half in an optimally trained WGAN. We have also proposed a new way\nof measuring GAN quality which is based on image completion tasks. We have\nevaluated our method on four visual datasets: CelebA, LSUN Bedroom, CIFAR-10,\nand 50K-SSFF, and have seen both visual and quantitative improvement over\nbaseline WGAN.","url_abs":"http://arxiv.org/abs/1704.04865v1","url_pdf":"http://arxiv.org/pdf/1704.04865v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"gang-of-gans-generative-adversarial-networks","repo_url":"https://github.com/human-analysis/RankGAN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[],"methods":[{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"wgan","method_name":"WGAN"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}