Few-Shot Learning as Domain Adaptation: Algorithm and Analysis
arXiv:2002.02050
Abstract
To recognize the unseen classes with only few samples, few-shot learning (FSL) uses prior knowledge learned from the seen classes. A major challenge for FSL is that the distribution of the unseen classes is different from that of those seen, resulting in poor generalization even when a model is meta-trained on the seen classes. This class-difference-caused distribution shift can be considered as a special case of domain shift. In this paper, for the first time, we propose a domain adaptation prototypical network with attention (DAPNA) to explicitly tackle such a domain shift problem in a meta-learning framework. Specifically, armed with a set transformer based attention module, we construct each episode with two sub-episodes without class overlap on the seen classes to simulate the domain shift between the seen and unseen classes. To align the feature distributions of the two sub-episodes with limited training samples, a feature transfer network is employed together with a margin disparity discrepancy (MDD) loss. Importantly, theoretical analysis is provided to give the learning bound of our DAPNA. Extensive experiments show that our DAPNA outperforms the state-of-the-art FSL alternatives, often by significant margins.
arXiv commentThere exist some mistakes in the experiments
References in corpus (15)
- Prototypical Networks for Few-shot Learning
- Learning Transferable Features with Deep Adaptation Networks
- Conditional Adversarial Domain Adaptation
- Meta-SGD: Learning to Learn Quickly for Few-Shot Learning
- Domain Generalization via Invariant Feature Representation
- Learning to Propagate Labels: Transductive Propagation Network for Few-shot Learning
- Domain Adaptation: Learning Bounds and Algorithms
- Learned in Translation: Contextualized Word Vectors
- Bridging Theory and Algorithm for Domain Adaptation
- Few-Shot Adversarial Domain Adaptation
- Meta-Learning with Latent Embedding Optimization
- Set Transformer: A Framework for Attention-based Permutation-Invariant Neural Networks
- Gradient-Based Meta-Learning with Learned Layerwise Metric and Subspace
- Dynamic Few-Shot Visual Learning without Forgetting
- Edge-labeling Graph Neural Network for Few-shot Learning