Learning to Learn Task Transformations for Improved Few-Shot Classification

被引:0
作者
Zheng, Guangtao [1 ]
Suo, Qiuling [2 ]
Huai, Mengdi [3 ]
Zhang, Aidong [1 ]
机构
[1] Univ Virginia, Dept Comp Sci, Charlottesville, VA 22904 USA
[2] Univ Buffalo, Dept Comp Sci & Engn, Buffalo, NY USA
[3] Iowa State Univ, Dept Comp Sci, Ames, IA USA
来源
PROCEEDINGS OF THE 2023 SIAM INTERNATIONAL CONFERENCE ON DATA MINING, SDM | 2023年
基金
美国国家科学基金会;
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
Meta-learning has shown great promise in few-shot image classification where only a small amount of labeled data is available in each classification task. Many training tasks are provided to train a meta-model that can quickly learn new and similar concepts with few labeled samples. Data augmentation is often used to augment training tasks to avoid overfitting. However, existing data augmentation methods are often manually designed and fixed during training, ignoring training dynamics and the di.erence between various meta-learning settings specified by meta-model architectures and meta-learning algorithms. To address this problem, we add a task transformation layer between a training task and a meta-model such that the right amount of perturbation is added to training tasks for a certain meta-learning setting at a certain training stage. By jointly optimizing the task transformation layer and the meta-model, we avoid the risk of providing tasks that are either too easy or too difficult during training. We design the task transformation layer as a stochastic transformation function, adding the flexibility in how a training task can be transformed. We leverage di.erentiable data augmentations as the building blocks of the task transformation function for efficient optimization. Extensive experiments show that our method can consistently improve the few-shot generalization performance of various meta-models trained with di.erent meta-learning algorithms, meta-model architectures, and datasets.
引用
收藏
页码:784 / 792
页数:9
相关论文
共 50 条
  • [21] Few-shot learning for short text classification
    Leiming Yan
    Yuhui Zheng
    Jie Cao
    [J]. Multimedia Tools and Applications, 2018, 77 : 29799 - 29810
  • [22] Dense Classification and Implanting for Few-Shot Learning
    Lifchitz, Yann
    Avrithis, Yannis
    Picard, Sylvaine
    Bursuc, Andrei
    [J]. 2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2019), 2019, : 9250 - 9259
  • [23] Diversified Contrastive Learning For Few-Shot Classification
    Lu, Guangtong
    Li, Fanzhang
    [J]. ARTIFICIAL NEURAL NETWORKS AND MACHINE LEARNING, ICANN 2023, PT I, 2023, 14254 : 147 - 158
  • [24] Integrative Few-Shot Learning for Classification and Segmentation
    Kang, Dahyun
    Cho, Minsu
    [J]. 2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2022, : 9969 - 9980
  • [25] Visual Classification of Malware by Few-shot Learning
    Tran, Kien
    Kubo, Masao
    Sato, Hiroshi
    [J]. PROCEEDINGS OF THE 2020 INTERNATIONAL CONFERENCE ON ARTIFICIAL LIFE AND ROBOTICS (ICAROB2020), 2020, : 770 - 774
  • [26] Few-Shot Learning for Medical Image Classification
    Cai, Aihua
    Hu, Wenxin
    Zheng, Jun
    [J]. ARTIFICIAL NEURAL NETWORKS AND MACHINE LEARNING, ICANN 2020, PT I, 2020, 12396 : 441 - 452
  • [27] Spatial Contrastive Learning for Few-Shot Classification
    Ouali, Yassine
    Hudelot, Celine
    Tami, Myriam
    [J]. MACHINE LEARNING AND KNOWLEDGE DISCOVERY IN DATABASES, 2021, 12975 : 671 - 686
  • [28] Continual Few-Shot Learning for Text Classification
    Pasunuru, Ramakanth
    Stoyanov, Veselin
    Bansal, Mohit
    [J]. 2021 CONFERENCE ON EMPIRICAL METHODS IN NATURAL LANGUAGE PROCESSING (EMNLP 2021), 2021, : 5688 - 5702
  • [29] NOISE SUPPRESSION FOR IMPROVED FEW-SHOT LEARNING
    Chen, Zhikui
    Ji, Tiandong
    Zhang, Suhua
    Zhong, Fangming
    [J]. 2022 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP), 2022, : 1900 - 1904
  • [30] FEW-SHOT CONTINUAL LEARNING FOR AUDIO CLASSIFICATION
    Wang, Yu
    Bryan, Nicholas J.
    Cartwright, Mark
    Bello, Juan Pablo
    Salamon, Justin
    [J]. 2021 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP 2021), 2021, : 321 - 325