{"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/stacked-quantizers-for-compositional-vector","title":"Stacked Quantizers for Compositional Vector Compression","arxiv_id":"1411.2173","date":"2014-11-08","proceeding":null,"authors":["Julieta Martinez","Holger H. Hoos","James J. Little"],"abstract":"Recently, Babenko and Lempitsky introduced Additive Quantization (AQ), a\ngeneralization of Product Quantization (PQ) where a non-independent set of\ncodebooks is used to compress vectors into small binary codes. Unfortunately,\nunder this scheme encoding cannot be done independently in each codebook, and\noptimal encoding is an NP-hard problem. In this paper, we observe that PQ and\nAQ are both compositional quantizers that lie on the extremes of the codebook\ndependence-independence assumption, and explore an intermediate approach that\nexploits a hierarchical structure in the codebooks. This results in a method\nthat achieves quantization error on par with or lower than AQ, while being\nseveral orders of magnitude faster. We perform a complexity analysis of PQ, AQ\nand our method, and evaluate our approach on standard benchmarks of SIFT and\nGIST descriptors, as well as on new datasets of features obtained from\nstate-of-the-art convolutional neural networks.","url_abs":"http://arxiv.org/abs/1411.2173v1","url_pdf":"http://arxiv.org/pdf/1411.2173v1.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":"stacked-quantizers-for-compositional-vector","repo_url":"https://github.com/una-dinosauria/Rayuela.jl","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}},{"paper_slug":"stacked-quantizers-for-compositional-vector","repo_url":"https://github.com/una-dinosauria/stacked-quantizers","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"quantization","task_name":"Quantization"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1411.2173","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}