Few-Shot Lifelong Learning

被引:0
|
作者
Mazumder, Pratik [1 ]
Singh, Pravendra
Rai, Piyush [1 ]
机构
[1] IIT Kanpur, Dept Comp Sci & Engn, Kanpur, Uttar Pradesh, India
来源
THIRTY-FIFTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, THIRTY-THIRD CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE AND THE ELEVENTH SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE | 2021年 / 35卷
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Many real-world classification problems often have classes with very few labeled training samples. Moreover, all possible classes may not be initially available for training, and may be given incrementally. Deep learning models need to deal with this two-fold problem in order to perform well in real-life situations. In this paper, we propose a novel Few-Shot Lifelong Learning (FSLL) method that enables deep learning models to perform lifelong/continual learning on few-shot data. Our method selects very few parameters from the model for training every new set of classes instead of training the full model. This helps in preventing overfitting. We choose the few parameters from the model in such a way that only the currently unimportant parameters get selected. By keeping the important parameters in the model intact, our approach minimizes catastrophic forgetting. Furthermore, we minimize the cosine similarity between the new and the old class prototypes in order to maximize their separation, thereby improving the classification performance. We also show that integrating our method with self-supervision improves the model performance significantly. We experimentally show that our method significantly outperforms existing methods on the miniImageNet, CIFAR-100, and CUB-200 datasets. Specifically, we outperform the state-of-the-art method by an absolute margin of 19.27% for the CUB dataset.
引用
收藏
页码:2337 / 2345
页数:9
相关论文
共 50 条
  • [11] Enhancing few-shot lifelong learning through fusion of cross-domain knowledge
    Zheng, knowledge Yaoyue
    Zhang, Xuetao
    Tian, Zhiqiang
    Du, Shaoyi
    INFORMATION FUSION, 2025, 115
  • [12] LFS-GAN: Lifelong Few-Shot Image Generation
    Seo, Juwon
    Kang, Ji-Su
    Park, Gyeong-Moon
    2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2023), 2023, : 11322 - 11332
  • [13] Co-Learning for Few-Shot Learning
    Rui Xu
    Lei Xing
    Shuai Shao
    Baodi Liu
    Kai Zhang
    Weifeng Liu
    Neural Processing Letters, 2022, 54 : 3339 - 3356
  • [14] RankDNN: Learning to Rank for Few-Shot Learning
    Guo, Qianyu
    Gong Haotong
    Wei, Xujun
    Fu, Yanwei
    Yu, Yizhou
    Zhang, Wenqiang
    Ge, Weifeng
    THIRTY-SEVENTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 37 NO 1, 2023, : 728 - 736
  • [15] Learning about few-shot concept learning
    Ananya Rastogi
    Nature Computational Science, 2022, 2 : 698 - 698
  • [16] Co-Learning for Few-Shot Learning
    Xu, Rui
    Xing, Lei
    Shao, Shuai
    Liu, Baodi
    Zhang, Kai
    Liu, Weifeng
    NEURAL PROCESSING LETTERS, 2022, 54 (04) : 3339 - 3356
  • [17] Federated Few-Shot Learning with Adversarial Learning
    Fan, Chenyou
    Huang, Jianwei
    2021 19TH INTERNATIONAL SYMPOSIUM ON MODELING AND OPTIMIZATION IN MOBILE, AD HOC, AND WIRELESS NETWORKS (WIOPT), 2021,
  • [18] Prototype Completion for Few-Shot Learning
    Zhang, Baoquan
    Li, Xutao
    Ye, Yunming
    Feng, Shanshan
    IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2023, 45 (10) : 12250 - 12268
  • [19] Few-Shot Learning With a Strong Teacher
    Ye, Han-Jia
    Ming, Lu
    Zhan, De-Chuan
    Chao, Wei-Lun
    IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2024, 46 (03) : 1425 - 1440
  • [20] Local Propagation for Few-Shot Learning
    Lifchitz, Yann
    Avrithis, Yannis
    Picard, Sylvaine
    2020 25TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION (ICPR), 2021, : 10457 - 10464