Meta-Reinforcement Learning in Non-Stationary and Dynamic Environments

In recent years, the subject of deep reinforcement learning (DRL) has developed very rapidly, and is now applied in various fields, such as decision making and control tasks. However, artificial agents trained with RL algorithms require great amounts of training data, unlike humans that are able to...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on pattern analysis and machine intelligence Vol. 45; no. 3; pp. 3476 - 3491
Main Authors: Bing, Zhenshan, Lerch, David, Huang, Kai, Knoll, Alois
Format: Journal Article
Language:English
Published: United States IEEE 01.03.2023
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Subjects:
ISSN:0162-8828, 1939-3539, 2160-9292, 1939-3539
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Abstract In recent years, the subject of deep reinforcement learning (DRL) has developed very rapidly, and is now applied in various fields, such as decision making and control tasks. However, artificial agents trained with RL algorithms require great amounts of training data, unlike humans that are able to learn new skills from very few examples. The concept of meta-reinforcement learning (meta-RL) has been recently proposed to enable agents to learn similar but new skills from a small amount of experience by leveraging a set of tasks with a shared structure. Due to the task representation learning strategy with few-shot adaptation, most recent work is limited to narrow task distributions and stationary environments, where tasks do not change within episodes. In this work, we address those limitations and introduce a training strategy that is applicable to non-stationary environments, as well as a task representation based on Gaussian mixture models to model clustered task distributions. We evaluate our method on several continuous robotic control benchmarks. Compared with state-of-the-art literature that is only applicable to stationary environments with few-shot adaption, our algorithm first achieves competitive asymptotic performance and superior sample efficiency in stationary environments with zero-shot adaption. Second, our algorithm learns to perform successfully in non-stationary settings as well as a continual learning setting, while learning well-structured task representations. Last, our algorithm learns basic distinct behaviors and well-structured task representations in task distributions with multiple qualitatively distinct tasks.
AbstractList In recent years, the subject of deep reinforcement learning (DRL) has developed very rapidly, and is now applied in various fields, such as decision making and control tasks. However, artificial agents trained with RL algorithms require great amounts of training data, unlike humans that are able to learn new skills from very few examples. The concept of meta-reinforcement learning (meta-RL) has been recently proposed to enable agents to learn similar but new skills from a small amount of experience by leveraging a set of tasks with a shared structure. Due to the task representation learning strategy with few-shot adaptation, most recent work is limited to narrow task distributions and stationary environments, where tasks do not change within episodes. In this work, we address those limitations and introduce a training strategy that is applicable to non-stationary environments, as well as a task representation based on Gaussian mixture models to model clustered task distributions. We evaluate our method on several continuous robotic control benchmarks. Compared with state-of-the-art literature that is only applicable to stationary environments with few-shot adaption, our algorithm first achieves competitive asymptotic performance and superior sample efficiency in stationary environments with zero-shot adaption. Second, our algorithm learns to perform successfully in non-stationary settings as well as a continual learning setting, while learning well-structured task representations. Last, our algorithm learns basic distinct behaviors and well-structured task representations in task distributions with multiple qualitatively distinct tasks.In recent years, the subject of deep reinforcement learning (DRL) has developed very rapidly, and is now applied in various fields, such as decision making and control tasks. However, artificial agents trained with RL algorithms require great amounts of training data, unlike humans that are able to learn new skills from very few examples. The concept of meta-reinforcement learning (meta-RL) has been recently proposed to enable agents to learn similar but new skills from a small amount of experience by leveraging a set of tasks with a shared structure. Due to the task representation learning strategy with few-shot adaptation, most recent work is limited to narrow task distributions and stationary environments, where tasks do not change within episodes. In this work, we address those limitations and introduce a training strategy that is applicable to non-stationary environments, as well as a task representation based on Gaussian mixture models to model clustered task distributions. We evaluate our method on several continuous robotic control benchmarks. Compared with state-of-the-art literature that is only applicable to stationary environments with few-shot adaption, our algorithm first achieves competitive asymptotic performance and superior sample efficiency in stationary environments with zero-shot adaption. Second, our algorithm learns to perform successfully in non-stationary settings as well as a continual learning setting, while learning well-structured task representations. Last, our algorithm learns basic distinct behaviors and well-structured task representations in task distributions with multiple qualitatively distinct tasks.
In recent years, the subject of deep reinforcement learning (DRL) has developed very rapidly, and is now applied in various fields, such as decision making and control tasks. However, artificial agents trained with RL algorithms require great amounts of training data, unlike humans that are able to learn new skills from very few examples. The concept of meta-reinforcement learning (meta-RL) has been recently proposed to enable agents to learn similar but new skills from a small amount of experience by leveraging a set of tasks with a shared structure. Due to the task representation learning strategy with few-shot adaptation, most recent work is limited to narrow task distributions and stationary environments, where tasks do not change within episodes. In this work, we address those limitations and introduce a training strategy that is applicable to non-stationary environments, as well as a task representation based on Gaussian mixture models to model clustered task distributions. We evaluate our method on several continuous robotic control benchmarks. Compared with state-of-the-art literature that is only applicable to stationary environments with few-shot adaption, our algorithm first achieves competitive asymptotic performance and superior sample efficiency in stationary environments with zero-shot adaption. Second, our algorithm learns to perform successfully in non-stationary settings as well as a continual learning setting, while learning well-structured task representations. Last, our algorithm learns basic distinct behaviors and well-structured task representations in task distributions with multiple qualitatively distinct tasks.
Author Lerch, David
Knoll, Alois
Bing, Zhenshan
Huang, Kai
Author_xml – sequence: 1
  givenname: Zhenshan
  orcidid: 0000-0002-0896-2517
  surname: Bing
  fullname: Bing, Zhenshan
  email: bing@in.tum.de
  organization: Department of Informatics, Technical University of Munich, München, Germany
– sequence: 2
  givenname: David
  surname: Lerch
  fullname: Lerch, David
  email: davidn.lerch@web.de
  organization: Department of Informatics, Technical University of Munich, München, Germany
– sequence: 3
  givenname: Kai
  orcidid: 0000-0003-0359-7810
  surname: Huang
  fullname: Huang, Kai
  email: huangk36@mail.sysu.edu.cn
  organization: Guangzhou and Peng cheng Laboratory, Sun Yat-sen University, Shenzhen, Guangdong, China
– sequence: 4
  givenname: Alois
  orcidid: 0000-0003-4840-076X
  surname: Knoll
  fullname: Knoll, Alois
  email: knoll@in.tum.de
  organization: Department of Informatics, Technical University of Munich, München, Germany
BackLink https://www.ncbi.nlm.nih.gov/pubmed/35737617$$D View this record in MEDLINE/PubMed
BookMark eNp9kcFOGzEQhi1EVQLtC4BUrcSFy6b22F7bRxQIIIW2aunZcnbHyChrU3uDxNuzaUIPOfQ0l-_7NTP_MTmMKSIhp4xOGaPm68OPy_u7KVCAKWdaSmEOyARYQ2sDBg7JhLIGaq1BH5HjUp4oZUJS_pEccam4apiakPk9Dq7-iSH6lFvsMQ7VAl2OIT5WIVbfUqx_DW4IKbr8WrnYVVev0fWhra7jS8gpbpTyiXzwblXw826ekN_z64fZbb34fnM3u1zULTdyqIXTDoVsqO88pY3qJAqFnRHggSolnIdlB2bc23DqPXDfNlx1nrUa9XKp-Am52OY-5_RnjWWwfSgtrlYuYloXC42mwNV4_4ie76FPaZ3juJ0FpbiQkjMxUl921HrZY2efc-jHQ-37h0ZAb4E2p1IyetuG7T-G7MLKMmo3Zdi_ZdhNGXZXxqjCnvqe_l_pbCsFRPwnGE2FAs3fAH4Dk8c
CODEN ITPIDJ
CitedBy_id crossref_primary_10_1016_j_phycom_2023_102054
crossref_primary_10_3389_fnbot_2024_1375309
crossref_primary_10_3389_fnbot_2023_1320251
crossref_primary_10_3389_fnbot_2024_1368243
crossref_primary_10_1109_TPAMI_2024_3386780
crossref_primary_10_3389_fmars_2023_1228306
crossref_primary_10_1109_TCYB_2025_3533035
crossref_primary_10_1109_TSC_2024_3451237
crossref_primary_10_1002_wene_70012
crossref_primary_10_3389_fnbot_2024_1362359
crossref_primary_10_1109_TKDE_2024_3382726
crossref_primary_10_1109_TCDS_2024_3485482
crossref_primary_10_3389_fnbot_2023_1277160
crossref_primary_10_1109_TKDE_2025_3535961
crossref_primary_10_1109_TCYB_2024_3356981
crossref_primary_10_1109_TCPMT_2024_3409310
crossref_primary_10_1109_TNNLS_2024_3385570
crossref_primary_10_1109_TPAMI_2024_3463709
Cites_doi 10.1109/IROS.2012.6386109
10.24963/ijcai.2019/387
ContentType Journal Article
Copyright Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2023
Copyright_xml – notice: Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2023
DBID 97E
RIA
RIE
AAYXX
CITATION
NPM
7SC
7SP
8FD
JQ2
L7M
L~C
L~D
7X8
DOI 10.1109/TPAMI.2022.3185549
DatabaseName IEEE Xplore (IEEE)
IEEE All-Society Periodicals Package (ASPP) 1998–Present
IEEE/IET Electronic Library (IEL)
CrossRef
PubMed
Computer and Information Systems Abstracts
Electronics & Communications Abstracts
Technology Research Database
ProQuest Computer Science Collection
Advanced Technologies Database with Aerospace
Computer and Information Systems Abstracts – Academic
Computer and Information Systems Abstracts Professional
MEDLINE - Academic
DatabaseTitle CrossRef
PubMed
Technology Research Database
Computer and Information Systems Abstracts – Academic
Electronics & Communications Abstracts
ProQuest Computer Science Collection
Computer and Information Systems Abstracts
Advanced Technologies Database with Aerospace
Computer and Information Systems Abstracts Professional
MEDLINE - Academic
DatabaseTitleList MEDLINE - Academic

PubMed
Technology Research Database
Database_xml – sequence: 1
  dbid: NPM
  name: PubMed
  url: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed
  sourceTypes: Index Database
– sequence: 2
  dbid: RIE
  name: IEEE/IET Electronic Library (IEL)
  url: https://ieeexplore.ieee.org/
  sourceTypes: Publisher
– sequence: 3
  dbid: 7X8
  name: MEDLINE - Academic
  url: https://search.proquest.com/medline
  sourceTypes: Aggregation Database
DeliveryMethod fulltext_linktorsrc
Discipline Engineering
Computer Science
EISSN 2160-9292
1939-3539
EndPage 3491
ExternalDocumentID 35737617
10_1109_TPAMI_2022_3185549
9804728
Genre orig-research
Journal Article
GrantInformation_xml – fundername: Key-Area Research and Development of Guangdong Province
  grantid: 2020B0101650001
– fundername: European Union's Horizon 2020 Framework Programme for Research and Innovation
  grantid: 945539
– fundername: Guangdong International cooperation program
  grantid: 2021A0505030024
GroupedDBID ---
-DZ
-~X
.DC
0R~
29I
4.4
53G
5GY
6IK
97E
AAJGR
AARMG
AASAJ
AAWTH
ABAZT
ABQJQ
ABVLG
ACGFO
ACGFS
ACIWK
ACNCT
AENEX
AGQYO
AHBIQ
AKJIK
AKQYR
ALMA_UNASSIGNED_HOLDINGS
ASUFR
ATWAV
BEFXN
BFFAM
BGNUA
BKEBE
BPEOZ
CS3
DU5
E.L
EBS
EJD
F5P
HZ~
IEDLZ
IFIPE
IPLJI
JAVBF
LAI
M43
MS~
O9-
OCL
P2P
PQQKQ
RIA
RIE
RNS
RXW
TAE
TN5
UHB
~02
5VS
9M8
AAYXX
ABFSI
ADRHT
AETEA
AETIX
AGSQL
AI.
AIBXA
ALLEH
CITATION
FA8
H~9
IBMZZ
ICLAB
IFJZH
RNI
RZB
VH1
AAYOK
NPM
PKN
RIC
RIG
XJT
Z5M
7SC
7SP
8FD
JQ2
L7M
L~C
L~D
7X8
ID FETCH-LOGICAL-c395t-4a8ae4560fdf0067d5e47ed942f20774af2bd29016930ff23fc637df1c8e8bb73
IEDL.DBID RIE
ISICitedReferencesCount 46
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000966202300001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 0162-8828
1939-3539
IngestDate Wed Oct 01 14:55:19 EDT 2025
Sun Jun 29 16:19:48 EDT 2025
Wed Feb 19 02:24:39 EST 2025
Sat Nov 29 02:58:20 EST 2025
Tue Nov 18 21:44:59 EST 2025
Wed Aug 27 02:47:54 EDT 2025
IsDoiOpenAccess false
IsOpenAccess true
IsPeerReviewed true
IsScholarly true
Issue 3
Language English
License https://ieeexplore.ieee.org/Xplorehelp/downloads/license-information/IEEE.html
https://doi.org/10.15223/policy-029
https://doi.org/10.15223/policy-037
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c395t-4a8ae4560fdf0067d5e47ed942f20774af2bd29016930ff23fc637df1c8e8bb73
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
ORCID 0000-0003-0359-7810
0000-0003-4840-076X
0000-0002-0896-2517
0000-0002-2348-6280
OpenAccessLink https://mediatum.ub.tum.de/1662422
PMID 35737617
PQID 2773455314
PQPubID 85458
PageCount 16
ParticipantIDs proquest_miscellaneous_2680237929
proquest_journals_2773455314
pubmed_primary_35737617
crossref_citationtrail_10_1109_TPAMI_2022_3185549
crossref_primary_10_1109_TPAMI_2022_3185549
ieee_primary_9804728
PublicationCentury 2000
PublicationDate 2023-03-01
PublicationDateYYYYMMDD 2023-03-01
PublicationDate_xml – month: 03
  year: 2023
  text: 2023-03-01
  day: 01
PublicationDecade 2020
PublicationPlace United States
PublicationPlace_xml – name: United States
– name: New York
PublicationTitle IEEE transactions on pattern analysis and machine intelligence
PublicationTitleAbbrev TPAMI
PublicationTitleAlternate IEEE Trans Pattern Anal Mach Intell
PublicationYear 2023
Publisher IEEE
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Publisher_xml – name: IEEE
– name: The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
References ref15
Zhang (ref27)
Ren (ref9)
Schmidhuber (ref2) 1987
Duan (ref3) 2016
Nagabandi (ref12) 2018
Rakelly (ref6) 2019
Wang (ref4) 2016
Mishra (ref13) 2017
Rothfuss (ref16) 2018
Humplik (ref17) 2019
Igl (ref26) 2018
Kingma (ref7) 2013
Gupta (ref23) 2018
Kingma (ref28) 2015
Vaswani (ref21) 2017
van den Oord (ref20) 2016
ref24
Nagabandi (ref11) 2018
Yu (ref14) 2019
(ref1) 2019
Wang (ref10) 2020
Brockman (ref18) 2016
Hausman (ref25)
Mnih (ref19)
Al-Shedivat (ref22) 2017
Finn (ref5) 2017
Haarnoja (ref8)
References_xml – start-page: 2850
  volume-title: Proc. 33rd Int. Conf. Mach. Learn.
  ident: ref19
  article-title: Asynchronous methods for deep reinforcement learning
– year: 2019
  ident: ref6
  article-title: Efficient off-policy meta-reinforcement learning via probabilistic context variables
– year: 2018
  ident: ref23
  article-title: Meta-reinforcement learning of structured exploration strategies
– year: 2019
  ident: ref1
  article-title: Solving Rubiks Cube with a robot hand
– year: 2016
  ident: ref20
  article-title: WaveNet: A generative model for raw audio
– year: 2018
  ident: ref26
  article-title: Deep variational reinforcement learning for POMDPs
– year: 2017
  ident: ref5
  article-title: Model-agnostic meta-learning for fast adaptation of deep networks
– year: 2015
  ident: ref28
  article-title: Variational dropout and the local reparameterization trick
– volume-title: Proc. 6th Int. Conf. Learn. Representations
  ident: ref25
  article-title: Learning an embedding space for transferable robot skills
– start-page: 2976
  volume-title: Proc. 35th Int. Conf. Mach. Learn.
  ident: ref8
  article-title: Soft actor-critic: Off-policy maximum entropy deep reinforcement learning with a stochastic actor
– year: 2016
  ident: ref18
  article-title: OpenAI Gym
– year: 1987
  ident: ref2
  article-title: Evolutionary principles in self-referential learning. (On learning how to learn: The meta hook.)
– year: 2018
  ident: ref11
  article-title: Learning to adapt in dynamic, real-world environments through meta-reinforcement learning
– year: 2017
  ident: ref13
  article-title: A simple neural attentive meta-learner
– year: 2018
  ident: ref16
  article-title: ProMP: Proximal meta-policy search
– year: 2013
  ident: ref7
  article-title: Auto-encoding variational bayes
– start-page: 12 767
  volume-title: Proc. 36th Int. Conf. Mach. Learn.
  ident: ref27
  article-title: LatentGNN: Learning efficient non-local relations for visual recognition
– year: 2019
  ident: ref17
  article-title: Meta reinforcement learning as task inference
– year: 2019
  ident: ref14
  article-title: Meta-world: A benchmark and evaluation for multi-task and meta reinforcement learning
– volume-title: Proc. Conf. Neural Inf. Process. Syst.
  ident: ref9
  article-title: Context-based meta-reinforcement learning with structured latent space
– year: 2017
  ident: ref21
  article-title: Attention is all you need
– year: 2016
  ident: ref4
  article-title: Learning to reinforcement learn
– year: 2018
  ident: ref12
  article-title: Deep online learning via meta-learning: Continual adaptation for model-based RL
– year: 2017
  ident: ref22
  article-title: Continuous adaptation via meta-learning in nonstationary and competitive environments
– ident: ref15
  doi: 10.1109/IROS.2012.6386109
– year: 2020
  ident: ref10
  article-title: Learning context-aware task reasoning for efficient meta-reinforcement learning
– year: 2016
  ident: ref3
  article-title: RL$^{2}$2: Fast reinforcement learning via slow reinforcement learning
– ident: ref24
  doi: 10.24963/ijcai.2019/387
SSID ssj0014503
Score 2.6250882
Snippet In recent years, the subject of deep reinforcement learning (DRL) has developed very rapidly, and is now applied in various fields, such as decision making and...
SourceID proquest
pubmed
crossref
ieee
SourceType Aggregation Database
Index Database
Enrichment Source
Publisher
StartPage 3476
SubjectTerms Adaptation models
Agents (artificial intelligence)
Algorithms
Control tasks
Decision making
Deep learning
Gaussian mixture model
Inference algorithms
Learning
Machine learning
Meta-reinforcement learning
Multitasking
Nonstationary environments
Probabilistic models
Representations
Robot control
robotic control
Robots
Skills
Strategy
task adaptation
Task analysis
task inference
Training
Title Meta-Reinforcement Learning in Non-Stationary and Dynamic Environments
URI https://ieeexplore.ieee.org/document/9804728
https://www.ncbi.nlm.nih.gov/pubmed/35737617
https://www.proquest.com/docview/2773455314
https://www.proquest.com/docview/2680237929
Volume 45
WOSCitedRecordID wos000966202300001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVIEE
  databaseName: IEEE/IET Electronic Library (IEL)
  customDbUrl:
  eissn: 2160-9292
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0014503
  issn: 0162-8828
  databaseCode: RIE
  dateStart: 19790101
  isFulltext: true
  titleUrlDefault: https://ieeexplore.ieee.org/
  providerName: IEEE
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3dSxwxEB9UStEHbbUfa1VS6FsbvcvHJnkU9WgfPKS1cG9LNpmIUPbKeSf0vzfJfuhDW_BtIdkPMjM7M5nM7wfwqUTvueU1FaXXVDBvaC2looaPg9UuxJShzmQTajrVs5m5WoMvQy8MIubDZ3icLnMt38_dKm2VnRidsA31OqwrpdperaFiIGRmQY4RTLTwmEb0DTIjc3J9dXr5LaaCjB2nXuGYEW3CSy5VtK3MU_bojzLByr9jzexzJjvP-9pXsN3FluS0VYbXsIbNLuz0vA2kM-Nd2HoCQrgHk0tcWvodM4Sqy7uFpENdvSG3DZnOG_qjLdjbxR9iG0_OWxp7cvGkS-4N_JxcXJ99pR27AnXcyCUVVluM4dMo-JB8lpcoFHojWGCjGBTawGqfqqyJLDEExoMrufJh7DTqulb8LWw08wbfA6nHiDbEsZgtCme5sU4Kq2SJyFAxLGDcr3HlOujxxIDxq8opyMhUWURVElHViaiAz8M9v1vgjf_O3ksCGGZ2a1_AQS_KqrPNu4opxYWM_x5RwMdhOFpVKpXYBuerOKdMwHgqxo4FvGtVYHh2rzn7f3_nB9hMlPTtObUD2FguVngIL9z98vZucRRVd6aPsuo-AF3m5-Y
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3dTxQxEJ8gGsEHURBcRa2Jb1K47cd2-0iUC0TuQvRMeNt026khIXvkuCPxv7ftfsiDmvi2Sbsf6czszHQ6vx_AhwKd44bXVBSupII5TWspFdU896a0PqQMdSKbUNNpeXmpL9bgYOiFQcR0-AwP42Wq5bu5XcWtsiNdRmzD8gE8lEKwvO3WGmoGQiYe5BDDBBsPiUTfIjPSR7OL48lZSAYZO4zdwiEn2oTHXKpgXYmp7LdHShQrf482k9cZb_3f9z6Dp110SY5bdXgOa9hsw1bP3EA6Q96GJ_dgCHdgPMGloV8xgajatF9IOtzVH-SqIdN5Q7-1JXuz-ElM48jnlsienNzrk3sB38cns0-ntONXoJZruaTClAZDADXyzkev5SQKhU4L5tkohIXGs9rFOmukS_SecW8LrpzPbYllXSu-C-vNvMGXQOoc0fgwFvJFYQ3XxkphlCwQGSqGGeT9Gle2Ax-PHBjXVUpCRrpKIqqiiKpORBl8HO65aaE3_jl7JwpgmNmtfQb7vSirzjpvK6YUFzL8fUQG74fhYFexWGIanK_CnCJC46kQPWaw16rA8Oxec179-Z3vYON0Njmvzs-mX17DZiSob0-t7cP6crHCN_DI3i2vbhdvkwL_Ah0l6kU
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Meta-Reinforcement+Learning+in+Non-Stationary+and+Dynamic+Environments&rft.jtitle=IEEE+transactions+on+pattern+analysis+and+machine+intelligence&rft.au=Bing%2C+Zhenshan&rft.au=Lerch%2C+David&rft.au=Huang%2C+Kai&rft.au=Knoll%2C+Alois&rft.date=2023-03-01&rft.pub=IEEE&rft.issn=0162-8828&rft.volume=45&rft.issue=3&rft.spage=3476&rft.epage=3491&rft_id=info:doi/10.1109%2FTPAMI.2022.3185549&rft_id=info%3Apmid%2F35737617&rft.externalDocID=9804728
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=0162-8828&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=0162-8828&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=0162-8828&client=summon