Rethinking Large-scale Pre-ranking System: Entire-chain Cross-domain Models
Industrial systems such as recommender systems and online advertising, have been widely equipped with multi-stage architectures, which are divided into several cascaded modules, including matching, pre-ranking, ranking and re-ranking. As a critical bridge between matching and ranking, existing pre-r...
Gespeichert in:
| Veröffentlicht in: | arXiv.org 2023-10 |
|---|---|
| Hauptverfasser: | , , , , , , , , , , , |
| Format: | Artikel |
| Sprache: | eng |
| Schlagworte: | |
| Online-Zugang: | Volltext |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| container_end_page | |
|---|---|
| container_issue | |
| container_start_page | |
| container_title | arXiv.org |
| container_volume | |
| creator | Song, Jinbo Huang, Ruoran Wang, Xinyang Huang, Wei Yu, Qian Chen, Mingming Yao, Yafei Fan, Chaosheng Peng, Changping Lin, Zhangang Hu, Jinghe Shao, Jingping |
| description | Industrial systems such as recommender systems and online advertising, have been widely equipped with multi-stage architectures, which are divided into several cascaded modules, including matching, pre-ranking, ranking and re-ranking. As a critical bridge between matching and ranking, existing pre-ranking approaches mainly endure sample selection bias (SSB) problem owing to ignoring the entire-chain data dependence, resulting in sub-optimal performances. In this paper, we rethink pre-ranking system from the perspective of the entire sample space, and propose Entire-chain Cross-domain Models (ECM), which leverage samples from the whole cascaded stages to effectively alleviate SSB problem. Besides, we design a fine-grained neural structure named ECMM to further improve the pre-ranking accuracy. Specifically, we propose a cross-domain multi-tower neural network to comprehensively predict for each stage result, and introduce the sub-networking routing strategy with \(L0\) regularization to reduce computational costs. Evaluations on real-world large-scale traffic logs demonstrate that our pre-ranking models outperform SOTA methods while time consumption is maintained within an acceptable level, which achieves better trade-off between efficiency and effectiveness. |
| doi_str_mv | 10.48550/arxiv.2310.08039 |
| format | Article |
| fullrecord | <record><control><sourceid>proquest_arxiv</sourceid><recordid>TN_cdi_arxiv_primary_2310_08039</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2876762186</sourcerecordid><originalsourceid>FETCH-LOGICAL-a526-75979093aaa6401295687a13c5532c0140cbe630ba4f44f51a46682e5e32c5a3</originalsourceid><addsrcrecordid>eNotj9tKw0AQhhdBsNQ-gFcGvN46e954J6EeMKJY78M03bSpOdTdVOzbu229Guabn-H_CLliMJVWKbhF_1v_TLmIACyI9IyMuBCMWsn5BZmEsAEArg1XSozIy4cb1nX3VXerJEe_cjSU2Ljk3Tvq8cTn-zC49i6ZdUMdcbnGuksy34dAl317WF77pWvCJTmvsAlu8j_HZP4w-8yeaP72-Jzd5xQV19So1KSQCkTUEhhPlbYGmShjH14Ck1AunBawQFlJWSmGUmvLnXLxrFCMyfXp61G02Pq6Rb8vDsLFUTgmbk6Jre-_dy4Mxabf-S5WKrg12mjOrBZ_FlBW6A</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2876762186</pqid></control><display><type>article</type><title>Rethinking Large-scale Pre-ranking System: Entire-chain Cross-domain Models</title><source>arXiv.org</source><source>Free E- Journals</source><creator>Song, Jinbo ; Huang, Ruoran ; Wang, Xinyang ; Huang, Wei ; Yu, Qian ; Chen, Mingming ; Yao, Yafei ; Fan, Chaosheng ; Peng, Changping ; Lin, Zhangang ; Hu, Jinghe ; Shao, Jingping</creator><creatorcontrib>Song, Jinbo ; Huang, Ruoran ; Wang, Xinyang ; Huang, Wei ; Yu, Qian ; Chen, Mingming ; Yao, Yafei ; Fan, Chaosheng ; Peng, Changping ; Lin, Zhangang ; Hu, Jinghe ; Shao, Jingping</creatorcontrib><description>Industrial systems such as recommender systems and online advertising, have been widely equipped with multi-stage architectures, which are divided into several cascaded modules, including matching, pre-ranking, ranking and re-ranking. As a critical bridge between matching and ranking, existing pre-ranking approaches mainly endure sample selection bias (SSB) problem owing to ignoring the entire-chain data dependence, resulting in sub-optimal performances. In this paper, we rethink pre-ranking system from the perspective of the entire sample space, and propose Entire-chain Cross-domain Models (ECM), which leverage samples from the whole cascaded stages to effectively alleviate SSB problem. Besides, we design a fine-grained neural structure named ECMM to further improve the pre-ranking accuracy. Specifically, we propose a cross-domain multi-tower neural network to comprehensively predict for each stage result, and introduce the sub-networking routing strategy with \(L0\) regularization to reduce computational costs. Evaluations on real-world large-scale traffic logs demonstrate that our pre-ranking models outperform SOTA methods while time consumption is maintained within an acceptable level, which achieves better trade-off between efficiency and effectiveness.</description><identifier>EISSN: 2331-8422</identifier><identifier>DOI: 10.48550/arxiv.2310.08039</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Computer Science - Artificial Intelligence ; Computer Science - Information Retrieval ; Computer Science - Learning ; Matching ; Neural networks ; Ranking ; Recommender systems ; Regularization</subject><ispartof>arXiv.org, 2023-10</ispartof><rights>2023. This work is published under http://arxiv.org/licenses/nonexclusive-distrib/1.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,780,881,27904</link.rule.ids><backlink>$$Uhttps://doi.org/10.1145/3511808.3557683$$DView published paper (Access to full text may be restricted)$$Hfree_for_read</backlink><backlink>$$Uhttps://doi.org/10.48550/arXiv.2310.08039$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Song, Jinbo</creatorcontrib><creatorcontrib>Huang, Ruoran</creatorcontrib><creatorcontrib>Wang, Xinyang</creatorcontrib><creatorcontrib>Huang, Wei</creatorcontrib><creatorcontrib>Yu, Qian</creatorcontrib><creatorcontrib>Chen, Mingming</creatorcontrib><creatorcontrib>Yao, Yafei</creatorcontrib><creatorcontrib>Fan, Chaosheng</creatorcontrib><creatorcontrib>Peng, Changping</creatorcontrib><creatorcontrib>Lin, Zhangang</creatorcontrib><creatorcontrib>Hu, Jinghe</creatorcontrib><creatorcontrib>Shao, Jingping</creatorcontrib><title>Rethinking Large-scale Pre-ranking System: Entire-chain Cross-domain Models</title><title>arXiv.org</title><description>Industrial systems such as recommender systems and online advertising, have been widely equipped with multi-stage architectures, which are divided into several cascaded modules, including matching, pre-ranking, ranking and re-ranking. As a critical bridge between matching and ranking, existing pre-ranking approaches mainly endure sample selection bias (SSB) problem owing to ignoring the entire-chain data dependence, resulting in sub-optimal performances. In this paper, we rethink pre-ranking system from the perspective of the entire sample space, and propose Entire-chain Cross-domain Models (ECM), which leverage samples from the whole cascaded stages to effectively alleviate SSB problem. Besides, we design a fine-grained neural structure named ECMM to further improve the pre-ranking accuracy. Specifically, we propose a cross-domain multi-tower neural network to comprehensively predict for each stage result, and introduce the sub-networking routing strategy with \(L0\) regularization to reduce computational costs. Evaluations on real-world large-scale traffic logs demonstrate that our pre-ranking models outperform SOTA methods while time consumption is maintained within an acceptable level, which achieves better trade-off between efficiency and effectiveness.</description><subject>Computer Science - Artificial Intelligence</subject><subject>Computer Science - Information Retrieval</subject><subject>Computer Science - Learning</subject><subject>Matching</subject><subject>Neural networks</subject><subject>Ranking</subject><subject>Recommender systems</subject><subject>Regularization</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>ABUWG</sourceid><sourceid>AFKRA</sourceid><sourceid>AZQEC</sourceid><sourceid>BENPR</sourceid><sourceid>CCPQU</sourceid><sourceid>DWQXO</sourceid><sourceid>GOX</sourceid><recordid>eNotj9tKw0AQhhdBsNQ-gFcGvN46e954J6EeMKJY78M03bSpOdTdVOzbu229Guabn-H_CLliMJVWKbhF_1v_TLmIACyI9IyMuBCMWsn5BZmEsAEArg1XSozIy4cb1nX3VXerJEe_cjSU2Ljk3Tvq8cTn-zC49i6ZdUMdcbnGuksy34dAl317WF77pWvCJTmvsAlu8j_HZP4w-8yeaP72-Jzd5xQV19So1KSQCkTUEhhPlbYGmShjH14Ck1AunBawQFlJWSmGUmvLnXLxrFCMyfXp61G02Pq6Rb8vDsLFUTgmbk6Jre-_dy4Mxabf-S5WKrg12mjOrBZ_FlBW6A</recordid><startdate>20231012</startdate><enddate>20231012</enddate><creator>Song, Jinbo</creator><creator>Huang, Ruoran</creator><creator>Wang, Xinyang</creator><creator>Huang, Wei</creator><creator>Yu, Qian</creator><creator>Chen, Mingming</creator><creator>Yao, Yafei</creator><creator>Fan, Chaosheng</creator><creator>Peng, Changping</creator><creator>Lin, Zhangang</creator><creator>Hu, Jinghe</creator><creator>Shao, Jingping</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20231012</creationdate><title>Rethinking Large-scale Pre-ranking System: Entire-chain Cross-domain Models</title><author>Song, Jinbo ; Huang, Ruoran ; Wang, Xinyang ; Huang, Wei ; Yu, Qian ; Chen, Mingming ; Yao, Yafei ; Fan, Chaosheng ; Peng, Changping ; Lin, Zhangang ; Hu, Jinghe ; Shao, Jingping</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a526-75979093aaa6401295687a13c5532c0140cbe630ba4f44f51a46682e5e32c5a3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Computer Science - Artificial Intelligence</topic><topic>Computer Science - Information Retrieval</topic><topic>Computer Science - Learning</topic><topic>Matching</topic><topic>Neural networks</topic><topic>Ranking</topic><topic>Recommender systems</topic><topic>Regularization</topic><toplevel>online_resources</toplevel><creatorcontrib>Song, Jinbo</creatorcontrib><creatorcontrib>Huang, Ruoran</creatorcontrib><creatorcontrib>Wang, Xinyang</creatorcontrib><creatorcontrib>Huang, Wei</creatorcontrib><creatorcontrib>Yu, Qian</creatorcontrib><creatorcontrib>Chen, Mingming</creatorcontrib><creatorcontrib>Yao, Yafei</creatorcontrib><creatorcontrib>Fan, Chaosheng</creatorcontrib><creatorcontrib>Peng, Changping</creatorcontrib><creatorcontrib>Lin, Zhangang</creatorcontrib><creatorcontrib>Hu, Jinghe</creatorcontrib><creatorcontrib>Shao, Jingping</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science & Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection><collection>arXiv Computer Science</collection><collection>arXiv.org</collection><jtitle>arXiv.org</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Song, Jinbo</au><au>Huang, Ruoran</au><au>Wang, Xinyang</au><au>Huang, Wei</au><au>Yu, Qian</au><au>Chen, Mingming</au><au>Yao, Yafei</au><au>Fan, Chaosheng</au><au>Peng, Changping</au><au>Lin, Zhangang</au><au>Hu, Jinghe</au><au>Shao, Jingping</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Rethinking Large-scale Pre-ranking System: Entire-chain Cross-domain Models</atitle><jtitle>arXiv.org</jtitle><date>2023-10-12</date><risdate>2023</risdate><eissn>2331-8422</eissn><abstract>Industrial systems such as recommender systems and online advertising, have been widely equipped with multi-stage architectures, which are divided into several cascaded modules, including matching, pre-ranking, ranking and re-ranking. As a critical bridge between matching and ranking, existing pre-ranking approaches mainly endure sample selection bias (SSB) problem owing to ignoring the entire-chain data dependence, resulting in sub-optimal performances. In this paper, we rethink pre-ranking system from the perspective of the entire sample space, and propose Entire-chain Cross-domain Models (ECM), which leverage samples from the whole cascaded stages to effectively alleviate SSB problem. Besides, we design a fine-grained neural structure named ECMM to further improve the pre-ranking accuracy. Specifically, we propose a cross-domain multi-tower neural network to comprehensively predict for each stage result, and introduce the sub-networking routing strategy with \(L0\) regularization to reduce computational costs. Evaluations on real-world large-scale traffic logs demonstrate that our pre-ranking models outperform SOTA methods while time consumption is maintained within an acceptable level, which achieves better trade-off between efficiency and effectiveness.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><doi>10.48550/arxiv.2310.08039</doi><oa>free_for_read</oa></addata></record> |
| fulltext | fulltext |
| identifier | EISSN: 2331-8422 |
| ispartof | arXiv.org, 2023-10 |
| issn | 2331-8422 |
| language | eng |
| recordid | cdi_arxiv_primary_2310_08039 |
| source | arXiv.org; Free E- Journals |
| subjects | Computer Science - Artificial Intelligence Computer Science - Information Retrieval Computer Science - Learning Matching Neural networks Ranking Recommender systems Regularization |
| title | Rethinking Large-scale Pre-ranking System: Entire-chain Cross-domain Models |
| url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-26T07%3A35%3A28IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_arxiv&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Rethinking%20Large-scale%20Pre-ranking%20System:%20Entire-chain%20Cross-domain%20Models&rft.jtitle=arXiv.org&rft.au=Song,%20Jinbo&rft.date=2023-10-12&rft.eissn=2331-8422&rft_id=info:doi/10.48550/arxiv.2310.08039&rft_dat=%3Cproquest_arxiv%3E2876762186%3C/proquest_arxiv%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2876762186&rft_id=info:pmid/&rfr_iscdi=true |