{"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/ae-ot-a-new-generative-model-based-on","title":"AE-OT: A NEW GENERATIVE MODEL BASED ON EXTENDED SEMI-DISCRETE OPTIMAL TRANSPORT","arxiv_id":null,"date":"2020-05-01","proceeding":"ICLR 2020 1","authors":["Dongsheng An","Yang Guo","Na lei","Zhongxuan Luo","Shing-Tung Yau","Xianfeng GU"],"abstract":"Generative adversarial networks (GANs) have attracted huge attention due to\nits capability to generate visual realistic images. However, most of the existing\nmodels suffer from the mode collapse or mode mixture problems. In this work, we\ngive a theoretic explanation of the both problems by Figalli’s regularity theory of\noptimal transportation maps. Basically, the generator compute the transportation\nmaps between the white noise distributions and the data distributions, which are\nin general discontinuous. However, DNNs can only represent continuous maps.\nThis intrinsic conflict induces mode collapse and mode mixture. In order to\ntackle the both problems, we explicitly separate the manifold embedding and the\noptimal transportation; the first part is carried out using an autoencoder to map the\nimages onto the latent space; the second part is accomplished using a GPU-based\nconvex optimization to find the discontinuous transportation maps. Composing the\nextended OT map and the decoder, we can finally generate new images from the\nwhite noise. This AE-OT model avoids representing discontinuous maps by DNNs,\ntherefore effectively prevents mode collapse and mode mixture.","url_abs":"https://openreview.net/forum?id=HkldyTNYwH","url_pdf":"https://openreview.net/pdf?id=HkldyTNYwH","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":"ae-ot-a-new-generative-model-based-on","repo_url":"https://github.com/k2cu8/pyOMT","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"decoder","task_name":"Decoder"},{"task_slug":null,"task_name":"GPU"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}