Rebalancing Multi-Label Class-Incremental Learning
Multi-label class-incremental learning (MLCIL) is essential for real-world multi-label applications, allowing models to learn new labels while retaining previously learned knowledge continuously. However, recent MLCIL approaches can only achieve suboptimal performance due to the oversight of the pos...
Saved in:
Published in | arXiv.org |
---|---|
Main Authors | , , , , , , , |
Format | Paper |
Language | English |
Published |
Ithaca
Cornell University Library, arXiv.org
22.08.2024
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Abstract | Multi-label class-incremental learning (MLCIL) is essential for real-world multi-label applications, allowing models to learn new labels while retaining previously learned knowledge continuously. However, recent MLCIL approaches can only achieve suboptimal performance due to the oversight of the positive-negative imbalance problem, which manifests at both the label and loss levels because of the task-level partial label issue. The imbalance at the label level arises from the substantial absence of negative labels, while the imbalance at the loss level stems from the asymmetric contributions of the positive and negative loss parts to the optimization. To address the issue above, we propose a Rebalance framework for both the Loss and Label levels (RebLL), which integrates two key modules: asymmetric knowledge distillation (AKD) and online relabeling (OR). AKD is proposed to rebalance at the loss level by emphasizing the negative label learning in classification loss and down-weighting the contribution of overconfident predictions in distillation loss. OR is designed for label rebalance, which restores the original class distribution in memory by online relabeling the missing classes. Our comprehensive experiments on the PASCAL VOC and MS-COCO datasets demonstrate that this rebalancing strategy significantly improves performance, achieving new state-of-the-art results even with a vanilla CNN backbone. |
---|---|
AbstractList | Multi-label class-incremental learning (MLCIL) is essential for real-world multi-label applications, allowing models to learn new labels while retaining previously learned knowledge continuously. However, recent MLCIL approaches can only achieve suboptimal performance due to the oversight of the positive-negative imbalance problem, which manifests at both the label and loss levels because of the task-level partial label issue. The imbalance at the label level arises from the substantial absence of negative labels, while the imbalance at the loss level stems from the asymmetric contributions of the positive and negative loss parts to the optimization. To address the issue above, we propose a Rebalance framework for both the Loss and Label levels (RebLL), which integrates two key modules: asymmetric knowledge distillation (AKD) and online relabeling (OR). AKD is proposed to rebalance at the loss level by emphasizing the negative label learning in classification loss and down-weighting the contribution of overconfident predictions in distillation loss. OR is designed for label rebalance, which restores the original class distribution in memory by online relabeling the missing classes. Our comprehensive experiments on the PASCAL VOC and MS-COCO datasets demonstrate that this rebalancing strategy significantly improves performance, achieving new state-of-the-art results even with a vanilla CNN backbone. |
Author | Shen, Yixi Xie, Junzhou Du, Kaile Liu, Guangcan Lyu, Fan Zhou, Yifan Li, Yuyang Hu, Fuyuan |
Author_xml | – sequence: 1 givenname: Kaile surname: Du fullname: Du, Kaile – sequence: 2 givenname: Yifan surname: Zhou fullname: Zhou, Yifan – sequence: 3 givenname: Fan surname: Lyu fullname: Lyu, Fan – sequence: 4 givenname: Yuyang surname: Li fullname: Li, Yuyang – sequence: 5 givenname: Junzhou surname: Xie fullname: Xie, Junzhou – sequence: 6 givenname: Yixi surname: Shen fullname: Shen, Yixi – sequence: 7 givenname: Fuyuan surname: Hu fullname: Hu, Fuyuan – sequence: 8 givenname: Guangcan surname: Liu fullname: Liu, Guangcan |
BookMark | eNqNys0KgkAUQOEhCrLyHYTWA-MdNVtLUWCbcC9XuYUyXWt-3j8XPUCrszjfRix5YlqICLROZZkBrEXs3KiUguIAea4jAXfq0CD3Az-TWzB-kDV2ZJLKoHPyyr2lF7FHk9SElme2E6sHGkfxr1uxP5-a6iLfdvoEcr4dp2B5Xq1WxyLTJaSl_k99AfOPNWA |
ContentType | Paper |
Copyright | 2024. 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. |
Copyright_xml | – notice: 2024. 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. |
DBID | 8FE 8FG ABJCF ABUWG AFKRA AZQEC BENPR BGLVJ CCPQU DWQXO HCIFZ L6V M7S PIMPY PQEST PQQKQ PQUKI PRINS PTHSS |
DatabaseName | ProQuest SciTech Collection ProQuest Technology Collection Materials Science & Engineering Collection ProQuest Central (Alumni) ProQuest Central ProQuest Central Essentials ProQuest Central Technology Collection ProQuest One Community College ProQuest Central Korea SciTech Premium Collection ProQuest Engineering Collection Engineering Database Publicly Available Content Database ProQuest One Academic Eastern Edition (DO NOT USE) ProQuest One Academic ProQuest One Academic UKI Edition ProQuest Central China Engineering Collection |
DatabaseTitle | Publicly Available Content Database Engineering Database Technology Collection ProQuest Central Essentials ProQuest One Academic Eastern Edition ProQuest Central (Alumni Edition) SciTech Premium Collection ProQuest One Community College ProQuest Technology Collection ProQuest SciTech Collection ProQuest Central China ProQuest Central ProQuest Engineering Collection ProQuest One Academic UKI Edition ProQuest Central Korea Materials Science & Engineering Collection ProQuest One Academic Engineering Collection |
DatabaseTitleList | Publicly Available Content Database |
Database_xml | – sequence: 1 dbid: 8FG name: ProQuest Technology Collection url: https://search.proquest.com/technologycollection1 sourceTypes: Aggregation Database |
DeliveryMethod | fulltext_linktorsrc |
Discipline | Physics |
EISSN | 2331-8422 |
Genre | Working Paper/Pre-Print |
GroupedDBID | 8FE 8FG ABJCF ABUWG AFKRA ALMA_UNASSIGNED_HOLDINGS AZQEC BENPR BGLVJ CCPQU DWQXO FRJ HCIFZ L6V M7S M~E PIMPY PQEST PQQKQ PQUKI PRINS PTHSS |
ID | FETCH-proquest_journals_30964382183 |
IEDL.DBID | 8FG |
IngestDate | Thu Oct 10 22:06:37 EDT 2024 |
IsOpenAccess | true |
IsPeerReviewed | false |
IsScholarly | false |
Language | English |
LinkModel | DirectLink |
MergedId | FETCHMERGED-proquest_journals_30964382183 |
OpenAccessLink | https://www.proquest.com/docview/3096438218?pq-origsite=%requestingapplication% |
PQID | 3096438218 |
PQPubID | 2050157 |
ParticipantIDs | proquest_journals_3096438218 |
PublicationCentury | 2000 |
PublicationDate | 20240822 |
PublicationDateYYYYMMDD | 2024-08-22 |
PublicationDate_xml | – month: 08 year: 2024 text: 20240822 day: 22 |
PublicationDecade | 2020 |
PublicationPlace | Ithaca |
PublicationPlace_xml | – name: Ithaca |
PublicationTitle | arXiv.org |
PublicationYear | 2024 |
Publisher | Cornell University Library, arXiv.org |
Publisher_xml | – name: Cornell University Library, arXiv.org |
SSID | ssj0002672553 |
Score | 3.5597436 |
SecondaryResourceType | preprint |
Snippet | Multi-label class-incremental learning (MLCIL) is essential for real-world multi-label applications, allowing models to learn new labels while retaining... |
SourceID | proquest |
SourceType | Aggregation Database |
SubjectTerms | Labels Learning Performance enhancement |
Title | Rebalancing Multi-Label Class-Incremental Learning |
URI | https://www.proquest.com/docview/3096438218 |
hasFullText | 1 |
inHoldings | 1 |
isFullTextHit | |
isPrint | |
link | http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwfV1NSwMxEB20i-CtfqG2lgW9Bncnu9n0JCi7FrGliEJvJclmvRSt3Xr1tzuJqR6EXgIhkC-SecnLywzAVYIqLWo0rNaJZBmahuk010wpQ2hOEKQT98F5PBGjl-xhls8C4dYGWeXGJnpDXb8bx5Ff88R5jpKESDfLD-aiRrnX1RBCYxeiFAvhJH2yuv_lWFAUdGLm_8ysx46qC9FULe3qAHbs2yHsecmlaY8AaWROV2gIPGL_D5Y9Km0XsY9TyWjj_lB3ahEHJ6ivx3BZlc93I7ZpaB6WQjv_6zg_gQ7d6e0pxEpmMm-yRgpKxVBqLmqZcBzatDAN5mfQ31bT-fbiHuwjYa-jPhH70FmvPu0FYedaD_wEDSC6LSfTJ8qNv8pvNoN4mQ |
link.rule.ids | 780,784,12765,21388,33373,33744,43600,43805 |
linkProvider | ProQuest |
linkToHtml | http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwfZ3PT8MgFMdfdIvRmz_jj6lN9EqkQCk7eTDWqt3iYSa7NUCpl0XnOv9_H8j0YLILFxJaGngf-Pb9ALimTKd5wyxpDFVEMNsSk2aGaG2R5oggQ32A82gsy1fxNM2mUXDrolvlyiYGQ918WK-R33DqM0cpJNLt_JP4qlH-72osobEJfcER3T5SvHj41ViYzPHEzP-Z2cCOYhf6L3ruFnuw4d73YSu4XNruABjOzPsVWoRHEuJgSaWNmyWhTiXBjfsj3elZEpOgvh3CVXE_uSvJ6kF1XApd_ffi_Ah6eKd3x5BoJVTWilZJbOVQGS4bRTkbujS3LctOYLBupNP13ZewXU5GVV09jp_PYIchh70MytgAesvFlztHji7NRfhY31EpeLA |
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=Rebalancing+Multi-Label+Class-Incremental+Learning&rft.jtitle=arXiv.org&rft.au=Du%2C+Kaile&rft.au=Zhou%2C+Yifan&rft.au=Lyu%2C+Fan&rft.au=Li%2C+Yuyang&rft.date=2024-08-22&rft.pub=Cornell+University+Library%2C+arXiv.org&rft.eissn=2331-8422 |