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
[31]   FEW-SHOT CONTINUAL LEARNING FOR AUDIO CLASSIFICATION [J].
Wang, Yu ;
Bryan, Nicholas J. ;
Cartwright, Mark ;
Bello, Juan Pablo ;
Salamon, Justin .
2021 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP 2021), 2021, :321-325
[32]   Spatial Contrastive Learning for Few-Shot Classification [J].
Ouali, Yassine ;
Hudelot, Celine ;
Tami, Myriam .
MACHINE LEARNING AND KNOWLEDGE DISCOVERY IN DATABASES, 2021, 12975 :671-686
[33]   Continual Few-Shot Learning for Text Classification [J].
Pasunuru, Ramakanth ;
Stoyanov, Veselin ;
Bansal, Mohit .
2021 CONFERENCE ON EMPIRICAL METHODS IN NATURAL LANGUAGE PROCESSING (EMNLP 2021), 2021, :5688-5702
[34]   Grad2Task: Improved Few-shot Text Classification Using Gradients for Task Representation [J].
Wang, Jixuan ;
Wang, Kuan-Chieh ;
Rudzicz, Frank ;
Brudno, Michael .
ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 34 (NEURIPS 2021), 2021, 34
[35]   Revisiting Local Descriptor for Improved Few-Shot Classification [J].
He, Jun ;
Hong, Richang ;
Liu, Xueliang ;
Xu, Mingliang ;
Sun, Qianru .
ACM TRANSACTIONS ON MULTIMEDIA COMPUTING COMMUNICATIONS AND APPLICATIONS, 2022, 18 (02)
[36]   Task Agnostic Meta-Learning for Few-Shot Learning [J].
Jamal, Muhammad Abdullah ;
Qi, Guo-Jun .
2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2019), 2019, :11711-11719
[37]   Task-Equivariant Graph Few-shot Learning [J].
Kim, Sungwon ;
Lee, Junseok ;
Lee, Namkyeong ;
Kim, Wonjoong ;
Choi, Seungyoon ;
Park, Chanyoung .
PROCEEDINGS OF THE 29TH ACM SIGKDD CONFERENCE ON KNOWLEDGE DISCOVERY AND DATA MINING, KDD 2023, 2023, :1120-1131
[38]   DETA: Denoised Task Adaptation for Few-Shot Learning [J].
Zhang, Ji ;
Gao, Lianli ;
Luo, Xu ;
Shen, Hengtao ;
Song, Jingkuan .
2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2023), 2023, :11507-11517
[39]   Task Encoding With Distribution Calibration for Few-Shot Learning [J].
Zhang, Jing ;
Zhang, Xinzhou ;
Wang, Zhe .
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, 2022, 32 (09) :6240-6252
[40]   Learn to aggregate global and local representations for few-shot learning [J].
Mounir Abdelaziz ;
Zuping Zhang .
Multimedia Tools and Applications, 2023, 82 :32991-33014