Rainbow Memory: Continual Learning with a Memory of Diverse Samples

Continual learning is a realistic learning scenario for AI models. Prevalent scenario of continual learning, however, assumes disjoint sets of classes as tasks and is less realistic rather artificial. Instead, we focus on 'blurry' task boundary; where tasks shares classes and is more reali...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Proceedings (IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Online) S. 8214 - 8223
Hauptverfasser: Bang, Jihwan, Kim, Heesu, Yoo, YoungJoon, Ha, Jung-Woo, Choi, Jonghyun
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 01.06.2021
Schlagworte:
Codes
Computer vision
Learning (artificial intelligence)
Memory management
Sampling methods
Training
Uncertainty
ISSN:1063-6919
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Abstract Continual learning is a realistic learning scenario for AI models. Prevalent scenario of continual learning, however, assumes disjoint sets of classes as tasks and is less realistic rather artificial. Instead, we focus on 'blurry' task boundary; where tasks shares classes and is more realistic and practical. To address such task, we argue the importance of diversity of samples in an episodic memory. To enhance the sample diversity in the memory, we propose a novel memory management strategy based on per-sample classification uncertainty and data augmentation, named Rainbow Memory (RM). With extensive empirical validations on MNIST, CIFAR10, CIFAR100, and ImageNet datasets, we show that the proposed method significantly improves the accuracy in blurry continual learning setups, outperforming state of the arts by large margins despite its simplicity. Code and data splits will be available in https://github.com/clovaai/rainbow-memory.
AbstractList Continual learning is a realistic learning scenario for AI models. Prevalent scenario of continual learning, however, assumes disjoint sets of classes as tasks and is less realistic rather artificial. Instead, we focus on 'blurry' task boundary; where tasks shares classes and is more realistic and practical. To address such task, we argue the importance of diversity of samples in an episodic memory. To enhance the sample diversity in the memory, we propose a novel memory management strategy based on per-sample classification uncertainty and data augmentation, named Rainbow Memory (RM). With extensive empirical validations on MNIST, CIFAR10, CIFAR100, and ImageNet datasets, we show that the proposed method significantly improves the accuracy in blurry continual learning setups, outperforming state of the arts by large margins despite its simplicity. Code and data splits will be available in https://github.com/clovaai/rainbow-memory.
Author Bang, Jihwan
Ha, Jung-Woo
Kim, Heesu
Yoo, YoungJoon
Choi, Jonghyun
Author_xml – sequence: 1
  givenname: Jihwan
  surname: Bang
  fullname: Bang, Jihwan
  email: jihwan.bang@navercorp.com
  organization: Search Solutions, Inc
– sequence: 2
  givenname: Heesu
  surname: Kim
  fullname: Kim, Heesu
  email: heesu.kim89@navercorp.com
  organization: NAVER CLOVA
– sequence: 3
  givenname: YoungJoon
  surname: Yoo
  fullname: Yoo, YoungJoon
  email: youngjoon.yoo@navercorp.com
  organization: NAVER CLOVA
– sequence: 4
  givenname: Jung-Woo
  surname: Ha
  fullname: Ha, Jung-Woo
  email: jungwoo.ha@navercorp.com
  organization: NAVER CLOVA
– sequence: 5
  givenname: Jonghyun
  surname: Choi
  fullname: Choi, Jonghyun
  email: jhc@gist.ac.kr
  organization: GIST
BookMark eNotzMlOwzAUQFGDQKIt_QJY-AcSnu14YodCGaQgUBm2lZ08g1HiVEmg6t-zoKu7ObpzcpL6hIRcMsgZA3tVfrysC1UInXPgLAcwjB-ROVNKFoUEy4_JjIESmbLMnpHlOH4DgOCMKWtmpFy7mHy_o0_Y9cP-mpZ9mmL6cS2t0A0ppk-6i9MXdQdB-0Bv4y8OI9JX121bHM_JaXDtiMtDF-T9bvVWPmTV8_1jeVNlkYOYMum9FwEb0RiJgmuNijnbWDS1gtpJgbXwDLUKCK4I0oK0PHgusAFQzokFufj_RkTcbIfYuWG_sVJrA0b8AViFTVY
CODEN IEEPAD
ContentType Conference Proceeding
DBID 6IE
6IH
CBEJK
RIE
RIO
DOI 10.1109/CVPR46437.2021.00812
DatabaseName IEEE Electronic Library (IEL) Conference Proceedings
IEEE Proceedings Order Plan (POP) 1998-present by volume
IEEE Xplore All Conference Proceedings
IEEE Electronic Library (IEL)
IEEE Proceedings Order Plans (POP) 1998-present
DatabaseTitleList
Database_xml – sequence: 1
  dbid: RIE
  name: IEEE Electronic Library (IEL)
  url: https://ieeexplore.ieee.org/
  sourceTypes: Publisher
DeliveryMethod fulltext_linktorsrc
Discipline Applied Sciences
EISBN 1665445092
9781665445092
EISSN 1063-6919
EndPage 8223
ExternalDocumentID 9577808
Genre orig-research
GrantInformation_xml – fundername: National Research Foundation
  funderid: 10.13039/501100001321
– fundername: Defense Acquisition Program Administration
  funderid: 10.13039/501100003626
GroupedDBID 6IE
6IH
6IL
6IN
AAWTH
ABLEC
ADZIZ
ALMA_UNASSIGNED_HOLDINGS
BEFXN
BFFAM
BGNUA
BKEBE
BPEOZ
CBEJK
CHZPO
IEGSK
IJVOP
OCL
RIE
RIL
RIO
ID FETCH-LOGICAL-i203t-5bbb3fed3d85e3277e61a9d9e8c60ca53ec3b1e76fe0a4f590592fb23ed006aa3
IEDL.DBID RIE
ISICitedReferencesCount 259
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000739917308044&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
IngestDate Wed Aug 27 02:28:31 EDT 2025
IsPeerReviewed false
IsScholarly true
Language English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-i203t-5bbb3fed3d85e3277e61a9d9e8c60ca53ec3b1e76fe0a4f590592fb23ed006aa3
PageCount 10
ParticipantIDs ieee_primary_9577808
PublicationCentury 2000
PublicationDate 2021-June
PublicationDateYYYYMMDD 2021-06-01
PublicationDate_xml – month: 06
  year: 2021
  text: 2021-June
PublicationDecade 2020
PublicationTitle Proceedings (IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Online)
PublicationTitleAbbrev CVPR
PublicationYear 2021
Publisher IEEE
Publisher_xml – name: IEEE
SSID ssj0003211698
Score 2.6365714
Snippet Continual learning is a realistic learning scenario for AI models. Prevalent scenario of continual learning, however, assumes disjoint sets of classes as tasks...
SourceID ieee
SourceType Publisher
StartPage 8214
SubjectTerms Codes
Computer vision
Learning (artificial intelligence)
Memory management
Sampling methods
Training
Uncertainty
Title Rainbow Memory: Continual Learning with a Memory of Diverse Samples
URI https://ieeexplore.ieee.org/document/9577808
WOSCitedRecordID wos000739917308044&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
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LSwMxEB5q8eCpaiu-ycGja3eT7CbxWi0etBRf9FbymEgvbelD8N-bbEP14MXThiUQmGSY5_cNwJWPdlg5nilrTcZ1yYLOoctMcG210N4ZXbPrP4rBQI5GatiA6y0WBhHr5jO8icu6lu9mdh1TZV1VCiEjsndHiGqD1drmU1iIZColEzquyFW39z585rEuFaJAWkQe0zh28tcMldqE9Fv_O3wfOj9YPDLcWpkDaOD0EFrJeSRJNZdt6KVKDXmKvbNftyTyTk0i4ShJHKofJCZdiU47yMyTu7orA8mLjiTByw689e9few9ZmpCQTWjOVllpjGEeHXOyREaFwKrQyimUtsptEDxaZgoUlcdcc1-q4ExRbyhDF7RNa3YEzelsisdACuQypxgels254EZSVTAZviGELY2UJ9COMhnPNyQY4ySO079_n8FeFPqmp-ocmqvFGi9g136uJsvFZX1z35USmoM
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LS8NAEB6KCnqq2opv9-DR2CSbdHe9VkvFthSt0lvZx0R6aaUPwX_vTrpUD148JYSFwMwO8_y-AbguyA8rl0XKWhNlOufe5tBFxoe2WujCGV2y63dFvy9HIzWowM0GC4OI5fAZ3tJr2ct3M7uiUllD5UJIQvZu0-asfI3W2lRUuM9lmkoGfFwSq0brbfCcUWfK54FpQkymtHjy1xaV0om0q__7_T7Uf9B4bLDxMwdQwekhVEP4yIJxLmrQCr0a1qPp2a87RsxTE6IcZYFF9Z1R2ZXpcILNCnZfzmUge9FEE7yow2v7YdjqRGFHQjRJY76McmMML9BxJ3PkqRDYTLRyCqVtxtaLHi03CYpmgbHOilz5cCotTMrReXvTmh_B1nQ2xWNgCWYyTtFfLRtnIjMyVQmX_umT2NxIeQI1ksn4Y02DMQ7iOP378xXsdoa97rj72H86gz1SwHrC6hy2lvMVXsCO_VxOFvPLUovf1xGdzg
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=proceeding&rft.title=Proceedings+%28IEEE+Computer+Society+Conference+on+Computer+Vision+and+Pattern+Recognition.+Online%29&rft.atitle=Rainbow+Memory%3A+Continual+Learning+with+a+Memory+of+Diverse+Samples&rft.au=Bang%2C+Jihwan&rft.au=Kim%2C+Heesu&rft.au=Yoo%2C+YoungJoon&rft.au=Ha%2C+Jung-Woo&rft.date=2021-06-01&rft.pub=IEEE&rft.eissn=1063-6919&rft.spage=8214&rft.epage=8223&rft_id=info:doi/10.1109%2FCVPR46437.2021.00812&rft.externalDocID=9577808