Large-Scale Few-Shot Learning: Knowledge Transfer With Class Hierarchy

By Aoxue Li, Tiange Luo, Zhiwu Lu, Tao A. Xiang, Liwei Wang. 2019

140 citations recorded by the source corpus.

Recently, large-scale few-shot learning (FSL) becomes topical. It is discovered that, for a large-scale FSL problem with 1,000 classes in the source domain, a strong baseline emerges, that is, simply training a deep feature embedding model using the aggregated source classes and performing nearest neighbor (NN) search using the learned features on the target classes. The state-of-the-art large-scale FSL methods struggle to beat this baseline, indicating intrinsic limitations on scalability. To overcome the challenge, we propose a novel large-scale FSL model by learning transferable visual features with the class hierarchy which encodes the semantic relations between source and target classes. Extensive experiments show that the proposed model significantly outperforms not only the NN baseline but also the state-of-the-art alternatives. Furthermore, we show that the proposed model can be easily extended to the large-scale zero-shot learning (ZSL) problem and also achieves the state-of-the-art results.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Incremental Few-Shot Object Detection (2020, neutral): “Note that there are many more recent FSL methods [6, 26, 7, 17, 47, 55, 50, 27, 39, 11, 16] that produce better performance on image classification tasks than MAML.”
Domain-Adaptive Few-Shot Learning (2020, background): “Unlike many visual recognition methods [17, 34, 60, 7, 1, 6, 8] that require a lot of supervision, few-shot learning (FSL) [9, 28, 23, 27, 16] aims to recognize a set of target”
Relational Generalized Few-Shot Learning (2019, background): “Alternatives to meta-learning include matching seen and novel feature spaces [38, 44], modeling the global class structure in the joint label space [24], obtaining transferable features from a hierarchy between seen and novel classes via clustering [21],”
Few-NERD: A Few-shot Named Entity Recognition Dataset (2021, background): “Knowledge transfer is crucial for few-shot learning (Li et al., 2019a).”
Few-Shot Learning With Global Class Representations (2019, background): “Inspired by the few-shot learning ability of humans, there has been increasing interest recently on few-shot learning (FSL) [8, 13, 23, 7, 25, 21, 18, 30, 24, 10].”
Compositional Few-Shot Recognition with Primitive Discovery and Enhancing (2020, methodology): “The baseline model refers to first conducting known-class classification with classic deep networks, then utilizing the network to extract feature for novel-class sample, and finally performing the Nearest Neighbor classification, which is widely used in current works [8, 23, 35, 46, 48, 51] and reg”
Boosting Few-Shot Learning With Adaptive Margin Loss (2020, methodology): “Then, the margin is integrated into the classiﬁcation loss to make similar classes more separable in the embedding space, which beneﬁts FSL. lem [6, 13, 25, 27].”
A Comprehensive Survey of Few-shot Learning: Evolution, Applications, Challenges, and Opportunities (2022, methodology): “[148] went further by using semantic information to model classes as hierarchy.”
Revisiting Mid-Level Patterns for Cross-Domain Few-Shot Recognition (2020, unknown): “Some works [21, 23, 24] in FSL utilize features of multiple appended layers to handle the hierarchy of classes.”