{"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/disentangled-representations-via-synergy","title":"Disentangled Representations via Synergy Minimization","arxiv_id":"1710.03839","date":"2017-10-10","proceeding":null,"authors":["Greg Ver Steeg","Rob Brekelmans","Hrayr Harutyunyan","Aram Galstyan"],"abstract":"Scientists often seek simplified representations of complex systems to\nfacilitate prediction and understanding. If the factors comprising a\nrepresentation allow us to make accurate predictions about our system, but\nobscuring any subset of the factors destroys our ability to make predictions,\nwe say that the representation exhibits informational synergy. We argue that\nsynergy is an undesirable feature in learned representations and that\nexplicitly minimizing synergy can help disentangle the true factors of\nvariation underlying data. We explore different ways of quantifying synergy,\nderiving new closed-form expressions in some cases, and then show how to modify\nlearning to produce representations that are minimally synergistic. We\nintroduce a benchmark task to disentangle separate characters from images of\nwords. We demonstrate that Minimally Synergistic (MinSyn) representations\ncorrectly disentangle characters while methods relying on statistical\nindependence fail.","url_abs":"http://arxiv.org/abs/1710.03839v1","url_pdf":"http://arxiv.org/pdf/1710.03839v1.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":"disentangled-representations-via-synergy","repo_url":"https://github.com/brekelma/character-disentanglement-emnist","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1710.03839","atlas_url":"https://app.syntology.ai/?focus=1710.03839","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}