{"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/deep-generative-models-for-distribution","title":"Deep Generative Models for Distribution-Preserving Lossy Compression","arxiv_id":"1805.11057","date":"2018-05-28","proceeding":"NeurIPS 2018 12","authors":["Michael Tschannen","Eirikur Agustsson","Mario Lucic"],"abstract":"We propose and study the problem of distribution-preserving lossy\ncompression. Motivated by recent advances in extreme image compression which\nallow to maintain artifact-free reconstructions even at very low bitrates, we\npropose to optimize the rate-distortion tradeoff under the constraint that the\nreconstructed samples follow the distribution of the training data. The\nresulting compression system recovers both ends of the spectrum: On one hand,\nat zero bitrate it learns a generative model of the data, and at high enough\nbitrates it achieves perfect reconstruction. Furthermore, for intermediate\nbitrates it smoothly interpolates between learning a generative model of the\ntraining data and perfectly reconstructing the training samples. We study\nseveral methods to approximately solve the proposed optimization problem,\nincluding a novel combination of Wasserstein GAN and Wasserstein Autoencoder,\nand present an extensive theoretical and empirical characterization of the\nproposed compression systems.","url_abs":"http://arxiv.org/abs/1805.11057v2","url_pdf":"http://arxiv.org/pdf/1805.11057v2.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":"deep-generative-models-for-distribution","repo_url":"https://github.com/mitscha/dplc","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"GPL-3.0"}}],"tasks":[{"task_slug":"image-compression","task_name":"Image Compression"},{"task_slug":"image-generation","task_name":"Image Generation"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1805.11057","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}