Online Continual Learning on Hierarchical Label Expansion

被引:0
|
作者
Lee, Byung Hyun [1 ]
Jung, Okchul [1 ]
Choi, Jonghyun [2 ,3 ]
Chun, Se Young [1 ,2 ]
机构
[1] Seoul Natl Univ, Dept ECE, Seoul, South Korea
[2] Seoul Natl Univ, INMC, Seoul, South Korea
[3] Yonsei Univ, Seoul, South Korea
基金
新加坡国家研究基金会;
关键词
D O I
10.1109/ICCV51070.2023.01080
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Continual learning (CL) enables models to adapt to new tasks and environments without forgetting previously learned knowledge. While current CL setups have ignored the relationship between labels in the past task and the new task with or without small task overlaps, real-world scenarios often involve hierarchical relationships between old and new tasks, posing another challenge for traditional CL approaches. To address this challenge, we propose a novel multi-level hierarchical class incremental task configuration with an online learning constraint, called hierarchical label expansion (HLE). Our configuration allows a network to first learn coarse-grained classes, with data labels continually expanding to more fine-grained classes in various hierarchy depths. To tackle this new setup, we propose a rehearsal-based method that utilizes hierarchy-aware pseudo-labeling to incorporate hierarchical class information. Additionally, we propose a simple yet effective memory management and sampling strategy that selectively adopts samples of newly encountered classes. Our experiments demonstrate that our proposed method can effectively use hierarchy on our HLE setup to improve classification accuracy across all levels of hierarchies, regardless of depth and class imbalance ratio, outperforming prior state-of-the-art works by significant margins while also outperforming them on the conventional disjoint, blurry and i-Blurry CL setups.
引用
收藏
页码:11727 / 11736
页数:10
相关论文
共 50 条
  • [1] Online Prototype Learning for Online Continual Learning
    Wei, Yujie
    Ye, Jiaxin
    Huang, Zhizhong
    Zhang, Junping
    Shan, Hongming
    2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2023), 2023, : 18718 - 18728
  • [2] Continual compression model for online continual learning
    Ye, Fei
    Bors, Adrian G.
    Applied Soft Computing, 2024, 167
  • [3] Hierarchical Correlations Replay for Continual Learning
    Wang, Qiang
    Liu, Jiayi
    Ji, Zhong
    Pang, Yanwei
    Zhang, Zhongfei
    KNOWLEDGE-BASED SYSTEMS, 2022, 250
  • [4] Learning on the Job: Online Lifelong and Continual Learning
    Liu, Bing
    THIRTY-FOURTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, THE THIRTY-SECOND INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE AND THE TENTH AAAI SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2020, 34 : 13544 - 13549
  • [5] ONLINE CONTINUAL LEARNING FOR EMBEDDED DEVICES
    Hayes, Tyler L.
    Kanan, Christopher
    CONFERENCE ON LIFELONG LEARNING AGENTS, VOL 199, 2022, 199
  • [6] Survey on Online Streaming Continual Learning
    Gunasekara, Nuwan
    Pfahringer, Bernhard
    Gomes, Heitor Murilo
    Bifet, Albert
    PROCEEDINGS OF THE THIRTY-SECOND INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, IJCAI 2023, 2023, : 6628 - 6637
  • [7] Sample Condensation in Online Continual Learning
    Sangermano, Mattia
    Carta, Antonio
    Cossu, Andrea
    Bacciu, Davide
    2022 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), 2022,
  • [8] Online continual learning with declarative memory
    Xiao, Zhe
    Du, Zhekai
    Wang, Ruijin
    Gan, Ruimeng
    Li, Jingjing
    NEURAL NETWORKS, 2023, 163 : 146 - 155
  • [9] Scalable Adversarial Online Continual Learning
    Dam, Tanmoy
    Pratama, Mahardhika
    Ferdaus, Meftahul
    Anavatti, Sreenatha
    Abbas, Hussein
    MACHINE LEARNING AND KNOWLEDGE DISCOVERY IN DATABASES, ECML PKDD 2022, PT III, 2023, 13715 : 373 - 389
  • [10] CarM: Hierarchical Episodic Memory for Continual Learning
    Lee, Soobee
    Weerakoon, Minindu
    Choi, Jonghyun
    Zhang, Minjia
    Wang, Di
    Jeon, Myeongjae
    PROCEEDINGS OF THE 59TH ACM/IEEE DESIGN AUTOMATION CONFERENCE, DAC 2022, 2022, : 1147 - 1152