Product quantization with dual codebooks for approximate nearest neighbor search. Pan, Z., Wang, L., Wang, Y., & Liu, Y. Neurocomputing, 401:59–68, August, 2020. ![Product quantization with dual codebooks for approximate nearest neighbor search [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper doi abstract bibtex Product quantization (PQ) is a powerful technique for approximate nearest neighbor (ANN) search. In this paper, to improve the accuracy of ANN search, we propose a new PQ-based method named product quantization with dual codebooks (DCPQ). Different from traditional PQ-based methods, we analyze quantization errors after learning the first PQ codebook, and then part of training vectors with larger quantization errors are found and selected to relearn a second PQ codebook. When encoding the database offline, all database vectors are firstly quantized using both of dual codebooks in each subspace, and the encoding mode of a database vector is determined after comparing the two quantization errors based on dual codebooks. Moreover, database vectors with the same encoding mode are grouped as a sub-database and can be more efficiently searched. Experimental results demonstrate that our proposed dual codebooks solution can achieve higher accuracy compared with the standard PQ and its variants.
@article{pan_product_2020,
title = {Product quantization with dual codebooks for approximate nearest neighbor search},
volume = {401},
issn = {0925-2312},
url = {https://www.sciencedirect.com/science/article/pii/S0925231220303519},
doi = {10.1016/j.neucom.2020.03.016},
abstract = {Product quantization (PQ) is a powerful technique for approximate nearest neighbor (ANN) search. In this paper, to improve the accuracy of ANN search, we propose a new PQ-based method named product quantization with dual codebooks (DCPQ). Different from traditional PQ-based methods, we analyze quantization errors after learning the first PQ codebook, and then part of training vectors with larger quantization errors are found and selected to relearn a second PQ codebook. When encoding the database offline, all database vectors are firstly quantized using both of dual codebooks in each subspace, and the encoding mode of a database vector is determined after comparing the two quantization errors based on dual codebooks. Moreover, database vectors with the same encoding mode are grouped as a sub-database and can be more efficiently searched. Experimental results demonstrate that our proposed dual codebooks solution can achieve higher accuracy compared with the standard PQ and its variants.},
language = {en},
urldate = {2023-08-13},
journal = {Neurocomputing},
author = {Pan, Zhibin and Wang, Liangzhuang and Wang, Yang and Liu, Yuchen},
month = aug,
year = {2020},
keywords = {\#Optimization, /unread, Approximate nearest neighbor search, Dual codebooks, Product quantization, Sub-database, Vector quantization},
pages = {59--68},
}
Downloads: 0
{"_id":"dfhibe85D443jx4qZ","bibbaseid":"pan-wang-wang-liu-productquantizationwithdualcodebooksforapproximatenearestneighborsearch-2020","author_short":["Pan, Z.","Wang, L.","Wang, Y.","Liu, Y."],"bibdata":{"bibtype":"article","type":"article","title":"Product quantization with dual codebooks for approximate nearest neighbor search","volume":"401","issn":"0925-2312","url":"https://www.sciencedirect.com/science/article/pii/S0925231220303519","doi":"10.1016/j.neucom.2020.03.016","abstract":"Product quantization (PQ) is a powerful technique for approximate nearest neighbor (ANN) search. In this paper, to improve the accuracy of ANN search, we propose a new PQ-based method named product quantization with dual codebooks (DCPQ). Different from traditional PQ-based methods, we analyze quantization errors after learning the first PQ codebook, and then part of training vectors with larger quantization errors are found and selected to relearn a second PQ codebook. When encoding the database offline, all database vectors are firstly quantized using both of dual codebooks in each subspace, and the encoding mode of a database vector is determined after comparing the two quantization errors based on dual codebooks. Moreover, database vectors with the same encoding mode are grouped as a sub-database and can be more efficiently searched. Experimental results demonstrate that our proposed dual codebooks solution can achieve higher accuracy compared with the standard PQ and its variants.","language":"en","urldate":"2023-08-13","journal":"Neurocomputing","author":[{"propositions":[],"lastnames":["Pan"],"firstnames":["Zhibin"],"suffixes":[]},{"propositions":[],"lastnames":["Wang"],"firstnames":["Liangzhuang"],"suffixes":[]},{"propositions":[],"lastnames":["Wang"],"firstnames":["Yang"],"suffixes":[]},{"propositions":[],"lastnames":["Liu"],"firstnames":["Yuchen"],"suffixes":[]}],"month":"August","year":"2020","keywords":"#Optimization, /unread, Approximate nearest neighbor search, Dual codebooks, Product quantization, Sub-database, Vector quantization","pages":"59–68","bibtex":"@article{pan_product_2020,\n\ttitle = {Product quantization with dual codebooks for approximate nearest neighbor search},\n\tvolume = {401},\n\tissn = {0925-2312},\n\turl = {https://www.sciencedirect.com/science/article/pii/S0925231220303519},\n\tdoi = {10.1016/j.neucom.2020.03.016},\n\tabstract = {Product quantization (PQ) is a powerful technique for approximate nearest neighbor (ANN) search. In this paper, to improve the accuracy of ANN search, we propose a new PQ-based method named product quantization with dual codebooks (DCPQ). Different from traditional PQ-based methods, we analyze quantization errors after learning the first PQ codebook, and then part of training vectors with larger quantization errors are found and selected to relearn a second PQ codebook. When encoding the database offline, all database vectors are firstly quantized using both of dual codebooks in each subspace, and the encoding mode of a database vector is determined after comparing the two quantization errors based on dual codebooks. Moreover, database vectors with the same encoding mode are grouped as a sub-database and can be more efficiently searched. Experimental results demonstrate that our proposed dual codebooks solution can achieve higher accuracy compared with the standard PQ and its variants.},\n\tlanguage = {en},\n\turldate = {2023-08-13},\n\tjournal = {Neurocomputing},\n\tauthor = {Pan, Zhibin and Wang, Liangzhuang and Wang, Yang and Liu, Yuchen},\n\tmonth = aug,\n\tyear = {2020},\n\tkeywords = {\\#Optimization, /unread, Approximate nearest neighbor search, Dual codebooks, Product quantization, Sub-database, Vector quantization},\n\tpages = {59--68},\n}\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n","author_short":["Pan, Z.","Wang, L.","Wang, Y.","Liu, Y."],"key":"pan_product_2020","id":"pan_product_2020","bibbaseid":"pan-wang-wang-liu-productquantizationwithdualcodebooksforapproximatenearestneighborsearch-2020","role":"author","urls":{"Paper":"https://www.sciencedirect.com/science/article/pii/S0925231220303519"},"keyword":["#Optimization","/unread","Approximate nearest neighbor search","Dual codebooks","Product quantization","Sub-database","Vector quantization"],"metadata":{"authorlinks":{}},"downloads":0,"html":""},"bibtype":"article","biburl":"https://bibbase.org/zotero/zzhenry2012","dataSources":["nZHrFJKyxKKDaWYM8"],"keywords":["#optimization","/unread","approximate nearest neighbor search","dual codebooks","product quantization","sub-database","vector quantization"],"search_terms":["product","quantization","dual","codebooks","approximate","nearest","neighbor","search","pan","wang","wang","liu"],"title":"Product quantization with dual codebooks for approximate nearest neighbor search","year":2020}