Continual Variational Autoencoder Learning via Online Cooperative Memorization

被引:8
|
作者
Ye, Fei [1 ]
Bors, Adrian G. [1 ]
机构
[1] Univ York, Dept Comp Sci, York YO10 5GH, N Yorkshire, England
来源
关键词
VAE; Continual learning; Lifelong generative modelling;
D O I
10.1007/978-3-031-20050-2_31
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Due to their inference, data representation and reconstruction properties, Variational Autoencoders (VAE) have been successfully used in continual learning classification tasks. However, their ability to generate images with specifications corresponding to the classes and databases learned during Continual Learning (CL) is not well understood and catastrophic forgetting remains a significant challenge. In this paper, we firstly analyze the forgetting behaviour of VAEs by developing a new theoretical framework that formulates CL as a dynamic optimal transport problem. This framework proves approximate bounds to the data likelihood without requiring the task information and explains how the prior knowledge is lost during the training process. We then propose a novel memory buffering approach, namely the Online Cooperative Memorization (OCM) framework, which consists of a Short-Term Memory (STM) that continually stores recent samples to provide future information for the model, and a Long-Term Memory (LTM) aiming to preserve a wide diversity of samples. The proposed OCM transfers certain samples from STM to LTM according to the information diversity selection criterion without requiring any supervised signals. The OCM framework is then combined with a dynamic VAE expansion mixture network for further enhancing its performance.
引用
收藏
页码:531 / 549
页数:19
相关论文
共 50 条
  • [1] Continual Variational Autoencoder via Continual Generative Knowledge Distillation
    Ye, Fei
    Bors, Adrian G.
    [J]. THIRTY-SEVENTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 37 NO 9, 2023, : 10918 - 10926
  • [2] CONTINUAL LEARNING FOR ANOMALY DETECTION WITH VARIATIONAL AUTOENCODER
    Wiewel, Felix
    Yang, Bin
    [J]. 2019 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP), 2019, : 3837 - 3841
  • [3] Wasserstein Expansible Variational Autoencoder for Discriminative and Generative Continual Learning
    Ye, Fei
    Bors, Adrian G.
    [J]. 2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2023), 2023, : 18619 - 18629
  • [4] Online Spatiotemporal Popularity Learning via Variational Bayes for Cooperative Caching
    Mehrizi, Sajad
    Chatterjee, Saikat
    Chatzinotas, Symeon
    Ottersten, Bjorn
    [J]. IEEE TRANSACTIONS ON COMMUNICATIONS, 2020, 68 (11) : 7068 - 7082
  • [5] Online Continual Learning Via Candidates Voting
    He, Jiangpeng
    Zhu, Fengqing
    [J]. 2022 IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV 2022), 2022, : 1292 - 1301
  • [6] CBA: Improving Online Continual Learning via Continual Bias Adaptor
    Wang, Quanziang
    Wang, Renzhen
    Wu, Yichen
    Jia, Xixi
    Meng, Deyu
    [J]. 2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2023), 2023, : 19036 - 19046
  • [7] Lifelong Variational Autoencoder via Online Adversarial Expansion Strategy
    Ye, Fei
    Bors, Adrian G.
    [J]. THIRTY-SEVENTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 37 NO 9, 2023, : 10909 - 10917
  • [8] Learning to sense from events via semantic variational autoencoder
    Silva Golo, Marcos Paulo
    Rossi, Rafael Geraldeli
    Marcacini, Ricardo Marcondes
    [J]. PLOS ONE, 2021, 16 (12):
  • [9] Online Prototype Learning for Online Continual Learning
    Wei, Yujie
    Ye, Jiaxin
    Huang, Zhizhong
    Zhang, Junping
    Shan, Hongming
    [J]. 2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2023), 2023, : 18718 - 18728
  • [10] Continual Referring Expression Comprehension via Dual Modular Memorization
    Shen, Heng Tao
    Chen, Cheng
    Wang, Peng
    Gao, Lianli
    Wang, Meng
    Song, Jingkuan
    [J]. IEEE TRANSACTIONS ON IMAGE PROCESSING, 2022, 31 : 6694 - 6706