A/B Testing and Best-arm Identification for Linear Bandits with Robustness to Non-stationarity
We investigate the fixed-budget best-arm identification (BAI) problem for linear bandits in a potentially non-stationary environment. Given a finite arm set $\mathcal{X}\subset\mathbb{R}^d$, a fixed budget $T$, and an unpredictable sequence of parameters $\left\lbrace\theta_t\right\rbrace_{t=1}^{T}$...
Gespeichert in:
Hauptverfasser: | , , , , |
---|---|
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
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 | Xiong, Zhihan Camilleri, Romain Fazel, Maryam Jain, Lalit Jamieson, Kevin |
description | We investigate the fixed-budget best-arm identification (BAI) problem for
linear bandits in a potentially non-stationary environment. Given a finite arm
set $\mathcal{X}\subset\mathbb{R}^d$, a fixed budget $T$, and an unpredictable
sequence of parameters $\left\lbrace\theta_t\right\rbrace_{t=1}^{T}$, an
algorithm will aim to correctly identify the best arm $x^* :=
\arg\max_{x\in\mathcal{X}}x^\top\sum_{t=1}^{T}\theta_t$ with probability as
high as possible. Prior work has addressed the stationary setting where
$\theta_t = \theta_1$ for all $t$ and demonstrated that the error probability
decreases as $\exp(-T /\rho^*)$ for a problem-dependent constant $\rho^*$. But
in many real-world $A/B/n$ multivariate testing scenarios that motivate our
work, the environment is non-stationary and an algorithm expecting a stationary
setting can easily fail. For robust identification, it is well-known that if
arms are chosen randomly and non-adaptively from a G-optimal design over
$\mathcal{X}$ at each time then the error probability decreases as
$\exp(-T\Delta^2_{(1)}/d)$, where $\Delta_{(1)} = \min_{x \neq x^*} (x^* -
x)^\top \frac{1}{T}\sum_{t=1}^T \theta_t$. As there exist environments where
$\Delta_{(1)}^2/ d \ll 1/ \rho^*$, we are motivated to propose a novel
algorithm $\mathsf{P1}$-$\mathsf{RAGE}$ that aims to obtain the best of both
worlds: robustness to non-stationarity and fast rates of identification in
benign settings. We characterize the error probability of
$\mathsf{P1}$-$\mathsf{RAGE}$ and demonstrate empirically that the algorithm
indeed never performs worse than G-optimal design but compares favorably to the
best algorithms in the stationary setting. |
doi_str_mv | 10.48550/arxiv.2307.15154 |
format | Article |
fullrecord | <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2307_15154</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2307_15154</sourcerecordid><originalsourceid>FETCH-LOGICAL-a674-e91a3182e51bebe8db0efe419f279f72bb3dd7d88b393f2a8881be2183cd85fe3</originalsourceid><addsrcrecordid>eNotj8tOwzAURL1hgQofwIr7A0ljOybOsql4VIqKhLImuq6vwRJ1kG0e_XtCYDWzODPSYeyKV2WtlarWGL_9Zylk1ZRccVWfs-fNuoOBUvbhBTBY6OZeYDzCzlLI3vkDZj8FcFOE3gfCCN3M-Zzgy-dXeJrMR8qBUoI8wX4KRcrLAqPPpwt25vAt0eV_rthwdztsH4r-8X633fQF3jR1QS1HybUgxQ0Z0tZU5KjmrRNN6xphjLS2sVob2UonUGs9g4JrebBaOZIrdv13uwiO79EfMZ7GX9FxEZU_9whPPw</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>A/B Testing and Best-arm Identification for Linear Bandits with Robustness to Non-stationarity</title><source>arXiv.org</source><creator>Xiong, Zhihan ; Camilleri, Romain ; Fazel, Maryam ; Jain, Lalit ; Jamieson, Kevin</creator><creatorcontrib>Xiong, Zhihan ; Camilleri, Romain ; Fazel, Maryam ; Jain, Lalit ; Jamieson, Kevin</creatorcontrib><description>We investigate the fixed-budget best-arm identification (BAI) problem for
linear bandits in a potentially non-stationary environment. Given a finite arm
set $\mathcal{X}\subset\mathbb{R}^d$, a fixed budget $T$, and an unpredictable
sequence of parameters $\left\lbrace\theta_t\right\rbrace_{t=1}^{T}$, an
algorithm will aim to correctly identify the best arm $x^* :=
\arg\max_{x\in\mathcal{X}}x^\top\sum_{t=1}^{T}\theta_t$ with probability as
high as possible. Prior work has addressed the stationary setting where
$\theta_t = \theta_1$ for all $t$ and demonstrated that the error probability
decreases as $\exp(-T /\rho^*)$ for a problem-dependent constant $\rho^*$. But
in many real-world $A/B/n$ multivariate testing scenarios that motivate our
work, the environment is non-stationary and an algorithm expecting a stationary
setting can easily fail. For robust identification, it is well-known that if
arms are chosen randomly and non-adaptively from a G-optimal design over
$\mathcal{X}$ at each time then the error probability decreases as
$\exp(-T\Delta^2_{(1)}/d)$, where $\Delta_{(1)} = \min_{x \neq x^*} (x^* -
x)^\top \frac{1}{T}\sum_{t=1}^T \theta_t$. As there exist environments where
$\Delta_{(1)}^2/ d \ll 1/ \rho^*$, we are motivated to propose a novel
algorithm $\mathsf{P1}$-$\mathsf{RAGE}$ that aims to obtain the best of both
worlds: robustness to non-stationarity and fast rates of identification in
benign settings. We characterize the error probability of
$\mathsf{P1}$-$\mathsf{RAGE}$ and demonstrate empirically that the algorithm
indeed never performs worse than G-optimal design but compares favorably to the
best algorithms in the stationary setting.</description><identifier>DOI: 10.48550/arxiv.2307.15154</identifier><language>eng</language><subject>Computer Science - Learning ; Statistics - Machine Learning</subject><creationdate>2023-07</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>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,881</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2307.15154$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2307.15154$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Xiong, Zhihan</creatorcontrib><creatorcontrib>Camilleri, Romain</creatorcontrib><creatorcontrib>Fazel, Maryam</creatorcontrib><creatorcontrib>Jain, Lalit</creatorcontrib><creatorcontrib>Jamieson, Kevin</creatorcontrib><title>A/B Testing and Best-arm Identification for Linear Bandits with Robustness to Non-stationarity</title><description>We investigate the fixed-budget best-arm identification (BAI) problem for
linear bandits in a potentially non-stationary environment. Given a finite arm
set $\mathcal{X}\subset\mathbb{R}^d$, a fixed budget $T$, and an unpredictable
sequence of parameters $\left\lbrace\theta_t\right\rbrace_{t=1}^{T}$, an
algorithm will aim to correctly identify the best arm $x^* :=
\arg\max_{x\in\mathcal{X}}x^\top\sum_{t=1}^{T}\theta_t$ with probability as
high as possible. Prior work has addressed the stationary setting where
$\theta_t = \theta_1$ for all $t$ and demonstrated that the error probability
decreases as $\exp(-T /\rho^*)$ for a problem-dependent constant $\rho^*$. But
in many real-world $A/B/n$ multivariate testing scenarios that motivate our
work, the environment is non-stationary and an algorithm expecting a stationary
setting can easily fail. For robust identification, it is well-known that if
arms are chosen randomly and non-adaptively from a G-optimal design over
$\mathcal{X}$ at each time then the error probability decreases as
$\exp(-T\Delta^2_{(1)}/d)$, where $\Delta_{(1)} = \min_{x \neq x^*} (x^* -
x)^\top \frac{1}{T}\sum_{t=1}^T \theta_t$. As there exist environments where
$\Delta_{(1)}^2/ d \ll 1/ \rho^*$, we are motivated to propose a novel
algorithm $\mathsf{P1}$-$\mathsf{RAGE}$ that aims to obtain the best of both
worlds: robustness to non-stationarity and fast rates of identification in
benign settings. We characterize the error probability of
$\mathsf{P1}$-$\mathsf{RAGE}$ and demonstrate empirically that the algorithm
indeed never performs worse than G-optimal design but compares favorably to the
best algorithms in the stationary setting.</description><subject>Computer Science - Learning</subject><subject>Statistics - Machine Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj8tOwzAURL1hgQofwIr7A0ljOybOsql4VIqKhLImuq6vwRJ1kG0e_XtCYDWzODPSYeyKV2WtlarWGL_9Zylk1ZRccVWfs-fNuoOBUvbhBTBY6OZeYDzCzlLI3vkDZj8FcFOE3gfCCN3M-Zzgy-dXeJrMR8qBUoI8wX4KRcrLAqPPpwt25vAt0eV_rthwdztsH4r-8X633fQF3jR1QS1HybUgxQ0Z0tZU5KjmrRNN6xphjLS2sVob2UonUGs9g4JrebBaOZIrdv13uwiO79EfMZ7GX9FxEZU_9whPPw</recordid><startdate>20230727</startdate><enddate>20230727</enddate><creator>Xiong, Zhihan</creator><creator>Camilleri, Romain</creator><creator>Fazel, Maryam</creator><creator>Jain, Lalit</creator><creator>Jamieson, Kevin</creator><scope>AKY</scope><scope>EPD</scope><scope>GOX</scope></search><sort><creationdate>20230727</creationdate><title>A/B Testing and Best-arm Identification for Linear Bandits with Robustness to Non-stationarity</title><author>Xiong, Zhihan ; Camilleri, Romain ; Fazel, Maryam ; Jain, Lalit ; Jamieson, Kevin</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a674-e91a3182e51bebe8db0efe419f279f72bb3dd7d88b393f2a8881be2183cd85fe3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Computer Science - Learning</topic><topic>Statistics - Machine Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Xiong, Zhihan</creatorcontrib><creatorcontrib>Camilleri, Romain</creatorcontrib><creatorcontrib>Fazel, Maryam</creatorcontrib><creatorcontrib>Jain, Lalit</creatorcontrib><creatorcontrib>Jamieson, Kevin</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>Xiong, Zhihan</au><au>Camilleri, Romain</au><au>Fazel, Maryam</au><au>Jain, Lalit</au><au>Jamieson, Kevin</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>A/B Testing and Best-arm Identification for Linear Bandits with Robustness to Non-stationarity</atitle><date>2023-07-27</date><risdate>2023</risdate><abstract>We investigate the fixed-budget best-arm identification (BAI) problem for
linear bandits in a potentially non-stationary environment. Given a finite arm
set $\mathcal{X}\subset\mathbb{R}^d$, a fixed budget $T$, and an unpredictable
sequence of parameters $\left\lbrace\theta_t\right\rbrace_{t=1}^{T}$, an
algorithm will aim to correctly identify the best arm $x^* :=
\arg\max_{x\in\mathcal{X}}x^\top\sum_{t=1}^{T}\theta_t$ with probability as
high as possible. Prior work has addressed the stationary setting where
$\theta_t = \theta_1$ for all $t$ and demonstrated that the error probability
decreases as $\exp(-T /\rho^*)$ for a problem-dependent constant $\rho^*$. But
in many real-world $A/B/n$ multivariate testing scenarios that motivate our
work, the environment is non-stationary and an algorithm expecting a stationary
setting can easily fail. For robust identification, it is well-known that if
arms are chosen randomly and non-adaptively from a G-optimal design over
$\mathcal{X}$ at each time then the error probability decreases as
$\exp(-T\Delta^2_{(1)}/d)$, where $\Delta_{(1)} = \min_{x \neq x^*} (x^* -
x)^\top \frac{1}{T}\sum_{t=1}^T \theta_t$. As there exist environments where
$\Delta_{(1)}^2/ d \ll 1/ \rho^*$, we are motivated to propose a novel
algorithm $\mathsf{P1}$-$\mathsf{RAGE}$ that aims to obtain the best of both
worlds: robustness to non-stationarity and fast rates of identification in
benign settings. We characterize the error probability of
$\mathsf{P1}$-$\mathsf{RAGE}$ and demonstrate empirically that the algorithm
indeed never performs worse than G-optimal design but compares favorably to the
best algorithms in the stationary setting.</abstract><doi>10.48550/arxiv.2307.15154</doi><oa>free_for_read</oa></addata></record> |
fulltext | fulltext_linktorsrc |
identifier | DOI: 10.48550/arxiv.2307.15154 |
ispartof | |
issn | |
language | eng |
recordid | cdi_arxiv_primary_2307_15154 |
source | arXiv.org |
subjects | Computer Science - Learning Statistics - Machine Learning |
title | A/B Testing and Best-arm Identification for Linear Bandits with Robustness to Non-stationarity |
url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-29T05%3A45%3A48IST&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=A/B%20Testing%20and%20Best-arm%20Identification%20for%20Linear%20Bandits%20with%20Robustness%20to%20Non-stationarity&rft.au=Xiong,%20Zhihan&rft.date=2023-07-27&rft_id=info:doi/10.48550/arxiv.2307.15154&rft_dat=%3Carxiv_GOX%3E2307_15154%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 |