Scheduled sampling for one-shot learning via matching network

• Published in: Pattern recognition 2019-12, Vol.96, p.106962, Article 106962
• Main Authors: Zhang, Lingling, Liu, Jun, Luo, Minnan, Chang, Xiaojun, Zheng, Qinghua, Hauptmann, Alexander G.
• Format: Article
• Language: English
• Subjects: Difficulty metric; From easy to difficult; Matching network; One-shot learning; Scheduled sampling
• ISSN: 0031-3203; 1873-5142
• DOI: 10.1016/j.patcog.2019.07.007

•A scheduled sampling strategy is introduced to adjust the training procedure of matching network, which accomplishes to learn the ability for one-shot prediction from easy to difficult.•We propose a novel metric to measure the difficulty of training samples, which jointly considers the diversity and similarity among the labels’ semantic. Samples with high-difficulty values are more difficult to learn for the matching network.•We conduct extensive experiments on datasets mini-Imagenet, Birds, and Flowers to illustrate the effectiveness and superiority of the proposed method. The experimental results demonstrate that our method consistently outperforms other competitors. Considering human can learn new object successfully from just one sample, one-shot learning, where each visual class just has one labeled sample for training, has attracted more and more attention. In the past years, most researchers achieve one-shot learning by training a matching network to map a small labeled support set and an unlabeled image to its label. The support set is combined by one image with the same label as unlabeled image and few images with other labels generated by random sampling. This random sampling strategy easily generates massive over-easy support sets in which most labels are less relevant to the label of unlabeled image. It leads to the limitation of matching network for one-shot prediction over indistinguishable label sets. For this issue, we propose a novel metric to evaluate the learning difficulty of support set, where this metric jointly considers the semantic diversity and similarity of visual labels. Based on the metric, we introduce a scheduled sampling strategy to train the matching network from easy to difficult. Extensive experimental results on three datasets, including mini-Imagenet, Birds and Flowers, indicate that our method could achieve significant improvements over other previous methods.