Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Hubs and Hyperspheres: Reducing Hubness and Improving Transductive Few-Shot Learning With Hyperspherical Embeddings
Daniel J. Trosten, Rwiddhi Chakraborty, Sigurd Løkse, Kristoffer Knutsen Wickstrøm, Robert Jenssen, Michael C. Kampffmeyer
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace EASE with noHub | SIAMESE with ResNet-18 features · mini-ImageNet 1-shot | accuracy: 71.79 -> 73.18 | Improved |
| replace EASE with noHub-S | SIAMESE with ResNet-18 features · tiered-ImageNet 1-shot | accuracy: 80.2 -> 81.34 | Improved |
| replace EASE with noHub-S | SIAMESE with ResNet-18 features · CUB-200 1-shot | accuracy: 81.88 -> 83.92 | Improved |
| replace noHub with noHub-S | SIAMESE with ResNet-18 features · mini-ImageNet 5-shot | accuracy: 82.58 -> 82.61 | No clear effect |
| replace noHub with noHub-S | SIAMESE with ResNet-18 features · CUB-200 5-shot | accuracy: 89.13 -> 88.93 | Hurt |
| add label guidance to L_align only (ablation) | noHub, tiered-ImageNet with ResNet-18 · 1-shot | accuracy: SimpleShot: 76.72 -> 78.25; 5-shot: 86.31 -> 85.46; SIAMESE: 82.94 -> 82.56; 5-shot: 87.87 -> 88.07 | Mixed |
| add label guidance to L_unif only (ablation) | noHub, tiered-ImageNet with ResNet-18 · 1-shot | accuracy: SimpleShot: 76.72 -> 78.33; 5-shot: 86.31 -> 86.15; SIAMESE: 82.94 -> 82.81; 5-shot: 87.87 -> 88.43 | Mixed |
| add label guidance to both L_align and L_unif (ablation) | noHub, tiered-ImageNet with ResNet-18 · 1-shot | accuracy: SimpleShot: 76.72 -> 78.35; SIAMESE: 82.94 -> 83.09 | Improved |
Closest papers by meaning
- Transductive Few-Shot Classification on the Oblique Manifold (ICCV 2021 · similarity 0.86)
- Learning To Affiliate: Mutual Centralized Learning for Few-Shot Classification (CVPR 2022 · similarity 0.86)
- Z-Score Normalization, Hubness, and Few-Shot Learning (ICCV 2021 · similarity 0.86)
- Deep Hyperspherical Learning (NeurIPS 2017 · similarity 0.85)
- Hyperbolic Image Embeddings (CVPR 2020 · similarity 0.85)
- DeepEMD: Few-Shot Image Classification With Differentiable Earth Mover's Distance and Structured Classifiers (CVPR 2020 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
