Hubs and Hyperspheres: Reducing Hubness and Improving Transductive Few-Shot Learning with Hyperspherical Embeddings
Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Documents
- Fulltext
Submitted manuscript, 1.39 MB, PDF document
Distance-based classification is frequently used in transductive few-shot learning (FSL). However, due to the high-dimensionality of image representations, FSL classifiers are prone to suffer from the hubness problem, where a few points (hubs) occur frequently in multiple nearest neighbour lists of other points. Hubness negatively impacts distance-based classification when hubs from one class appear often among the nearest neighbors of points from another class, degrading the classifier's performance. To address the hubness problem in FSL, we first prove that hubness can be eliminated by distributing representations uniformly on the hypersphere. We then propose two new approaches to embed representations on the hypersphere, which we prove optimize a tradeoff between uniformity and local similarity preservation - reducing hubness while retaining class structure. Our experiments show that the proposed methods reduce hubness, and significantly improves transductive FSL accuracy for a wide range of classifiers 11Code available at https://github.com/uitml/noHub..
Original language | English |
---|---|
Title of host publication | Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023 |
Number of pages | 10 |
Publisher | IEEE Computer Society Press |
Publication date | 2023 |
Pages | 7527-7536 |
ISBN (Electronic) | 9798350301298 |
DOIs | |
Publication status | Published - 2023 |
Event | 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023 - Vancouver, Canada Duration: 18 Jun 2023 → 22 Jun 2023 |
Conference
Conference | 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023 |
---|---|
Land | Canada |
By | Vancouver |
Periode | 18/06/2023 → 22/06/2023 |
Sponsor | Amazon Science, Ant Research, Cruise, et al., Google, Lambda |
Bibliographical note
Publisher Copyright:
© 2023 IEEE.
- continual, low-shot, meta, or long-tail learning, Transfer
Research areas
ID: 371288564