{"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/composite-quantization","title":"Composite Quantization","arxiv_id":"1712.00955","date":"2017-12-04","proceeding":null,"authors":["Jingdong Wang","Ting Zhang"],"abstract":"This paper studies the compact coding approach to approximate nearest\nneighbor search. We introduce a composite quantization framework. It uses the\ncomposition of several ($M$) elements, each of which is selected from a\ndifferent dictionary, to accurately approximate a $D$-dimensional vector, thus\nyielding accurate search, and represents the data vector by a short code\ncomposed of the indices of the selected elements in the corresponding\ndictionaries. Our key contribution lies in introducing a near-orthogonality\nconstraint, which makes the search efficiency is guaranteed as the cost of the\ndistance computation is reduced to $O(M)$ from $O(D)$ through a distance table\nlookup scheme. The resulting approach is called near-orthogonal composite\nquantization. We theoretically justify the equivalence between near-orthogonal\ncomposite quantization and minimizing an upper bound of a function formed by\njointly considering the quantization error and the search cost according to a\ngeneralized triangle inequality. We empirically show the efficacy of the\nproposed approach over several benchmark datasets. In addition, we demonstrate\nthe superior performances in other three applications: combination with\ninverted multi-index, quantizing the query for mobile search, and inner-product\nsimilarity search.","url_abs":"http://arxiv.org/abs/1712.00955v1","url_pdf":"http://arxiv.org/pdf/1712.00955v1.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":"composite-quantization","repo_url":"https://github.com/una-dinosauria/Rayuela.jl","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"quantization","task_name":"Quantization"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1712.00955","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}