Cascading Bandits: Learning to Rank in the Cascade Model

Cascading Bandits: Learning to Rank in the Cascade Model

A search engine usually outputs a list of $K$ web pages. The user examines this list, from the first web page to the last, and chooses the first attractive page. This model of user behavior is known as the cascade model. In this paper, we propose cascading bandits, a learning variant of the cascade...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Kveton, Branislav, Szepesvari, Csaba, Wen, Zheng, Ashkan, Azin
Format: Artikel
Sprache:eng
Schlagworte:
Computer Science - Learning
Statistics - Machine Learning
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Kveton, Branislav
Szepesvari, Csaba
Wen, Zheng
Ashkan, Azin
description A search engine usually outputs a list of $K$ web pages. The user examines this list, from the first web page to the last, and chooses the first attractive page. This model of user behavior is known as the cascade model. In this paper, we propose cascading bandits, a learning variant of the cascade model where the objective is to identify $K$ most attractive items. We formulate our problem as a stochastic combinatorial partial monitoring problem. We propose two algorithms for solving it, CascadeUCB1 and CascadeKL-UCB. We also prove gap-dependent upper bounds on the regret of these algorithms and derive a lower bound on the regret in cascading bandits. The lower bound matches the upper bound of CascadeKL-UCB up to a logarithmic factor. We experiment with our algorithms on several problems. The algorithms perform surprisingly well even when our modeling assumptions are violated.
doi_str_mv 10.48550/arxiv.1502.02763
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_1502_02763</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>1502_02763</sourcerecordid><originalsourceid>FETCH-LOGICAL-a673-8d916ca3ebda9e403b6283c55a28bd4ca02df7c7b6475d1a5af4adeb00488cf83</originalsourceid><addsrcrecordid>eNotj8mKwkAYhPsyB8n4AJ6mXyCx07tzG4MbRITBe_h7iTaT6UgSRN_e9VRQFFX1ITTJSca1EGQK3SWcs1wQmhGqJBshXUBvwYV4wHOILgz9Ny49dPHhDC3-hfiHQ8TD0eNX1ONt63zziT5qaHo_fmuC9svFvlin5W61KX7KFKRiqXazXFpg3jiYeU6YkVQzKwRQbRy3QKirlVVGciVcDgJqfp8whHCtba1Zgr5etc_r1akL_9BdqwdC9URgNwHiQGM</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Cascading Bandits: Learning to Rank in the Cascade Model</title><source>arXiv.org</source><creator>Kveton, Branislav ; Szepesvari, Csaba ; Wen, Zheng ; Ashkan, Azin</creator><creatorcontrib>Kveton, Branislav ; Szepesvari, Csaba ; Wen, Zheng ; Ashkan, Azin</creatorcontrib><description>A search engine usually outputs a list of $K$ web pages. The user examines this list, from the first web page to the last, and chooses the first attractive page. This model of user behavior is known as the cascade model. In this paper, we propose cascading bandits, a learning variant of the cascade model where the objective is to identify $K$ most attractive items. We formulate our problem as a stochastic combinatorial partial monitoring problem. We propose two algorithms for solving it, CascadeUCB1 and CascadeKL-UCB. We also prove gap-dependent upper bounds on the regret of these algorithms and derive a lower bound on the regret in cascading bandits. The lower bound matches the upper bound of CascadeKL-UCB up to a logarithmic factor. We experiment with our algorithms on several problems. The algorithms perform surprisingly well even when our modeling assumptions are violated.</description><identifier>DOI: 10.48550/arxiv.1502.02763</identifier><language>eng</language><subject>Computer Science - Learning ; Statistics - Machine Learning</subject><creationdate>2015-02</creationdate><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>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,780,885</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/1502.02763$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.1502.02763$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Kveton, Branislav</creatorcontrib><creatorcontrib>Szepesvari, Csaba</creatorcontrib><creatorcontrib>Wen, Zheng</creatorcontrib><creatorcontrib>Ashkan, Azin</creatorcontrib><title>Cascading Bandits: Learning to Rank in the Cascade Model</title><description>A search engine usually outputs a list of $K$ web pages. The user examines this list, from the first web page to the last, and chooses the first attractive page. This model of user behavior is known as the cascade model. In this paper, we propose cascading bandits, a learning variant of the cascade model where the objective is to identify $K$ most attractive items. We formulate our problem as a stochastic combinatorial partial monitoring problem. We propose two algorithms for solving it, CascadeUCB1 and CascadeKL-UCB. We also prove gap-dependent upper bounds on the regret of these algorithms and derive a lower bound on the regret in cascading bandits. The lower bound matches the upper bound of CascadeKL-UCB up to a logarithmic factor. We experiment with our algorithms on several problems. The algorithms perform surprisingly well even when our modeling assumptions are violated.</description><subject>Computer Science - Learning</subject><subject>Statistics - Machine Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2015</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj8mKwkAYhPsyB8n4AJ6mXyCx07tzG4MbRITBe_h7iTaT6UgSRN_e9VRQFFX1ITTJSca1EGQK3SWcs1wQmhGqJBshXUBvwYV4wHOILgz9Ny49dPHhDC3-hfiHQ8TD0eNX1ONt63zziT5qaHo_fmuC9svFvlin5W61KX7KFKRiqXazXFpg3jiYeU6YkVQzKwRQbRy3QKirlVVGciVcDgJqfp8whHCtba1Zgr5etc_r1akL_9BdqwdC9URgNwHiQGM</recordid><startdate>20150209</startdate><enddate>20150209</enddate><creator>Kveton, Branislav</creator><creator>Szepesvari, Csaba</creator><creator>Wen, Zheng</creator><creator>Ashkan, Azin</creator><scope>AKY</scope><scope>EPD</scope><scope>GOX</scope></search><sort><creationdate>20150209</creationdate><title>Cascading Bandits: Learning to Rank in the Cascade Model</title><author>Kveton, Branislav ; Szepesvari, Csaba ; Wen, Zheng ; Ashkan, Azin</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a673-8d916ca3ebda9e403b6283c55a28bd4ca02df7c7b6475d1a5af4adeb00488cf83</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2015</creationdate><topic>Computer Science - Learning</topic><topic>Statistics - Machine Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Kveton, Branislav</creatorcontrib><creatorcontrib>Szepesvari, Csaba</creatorcontrib><creatorcontrib>Wen, Zheng</creatorcontrib><creatorcontrib>Ashkan, Azin</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv Statistics</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Kveton, Branislav</au><au>Szepesvari, Csaba</au><au>Wen, Zheng</au><au>Ashkan, Azin</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Cascading Bandits: Learning to Rank in the Cascade Model</atitle><date>2015-02-09</date><risdate>2015</risdate><abstract>A search engine usually outputs a list of $K$ web pages. The user examines this list, from the first web page to the last, and chooses the first attractive page. This model of user behavior is known as the cascade model. In this paper, we propose cascading bandits, a learning variant of the cascade model where the objective is to identify $K$ most attractive items. We formulate our problem as a stochastic combinatorial partial monitoring problem. We propose two algorithms for solving it, CascadeUCB1 and CascadeKL-UCB. We also prove gap-dependent upper bounds on the regret of these algorithms and derive a lower bound on the regret in cascading bandits. The lower bound matches the upper bound of CascadeKL-UCB up to a logarithmic factor. We experiment with our algorithms on several problems. The algorithms perform surprisingly well even when our modeling assumptions are violated.</abstract><doi>10.48550/arxiv.1502.02763</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.1502.02763
ispartof
issn
language eng
recordid cdi_arxiv_primary_1502_02763
source arXiv.org
subjects Computer Science - Learning
Statistics - Machine Learning
title Cascading Bandits: Learning to Rank in the Cascade Model
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-26T05%3A23%3A13IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Cascading%20Bandits:%20Learning%20to%20Rank%20in%20the%20Cascade%20Model&rft.au=Kveton,%20Branislav&rft.date=2015-02-09&rft_id=info:doi/10.48550/arxiv.1502.02763&rft_dat=%3Carxiv_GOX%3E1502_02763%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true