Research map / Label-efficient and robust learning / Recognition and long-tail
ACE: Ally Complementary Experts for Solving Long-Tailed Recognition in One-Shot
Jiarui Cai, Yizhou Wang, Jenq-Neng Hwang
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 |
|---|---|---|---|
| add the complementary loss L_com (ablation) | ACE (3 experts), CIFAR100-LT IF=100 · CIFAR100-LT IF=100 | top-1 accuracy: 47.2 -> 49.4 | Improved |
| add the complementary loss L_com (many-shot) (ablation) | ACE (3 experts), CIFAR100-LT IF=100 · CIFAR100-LT IF=100 | many-shot accuracy: 71.5 -> 66.1 | Hurt |
| add the complementary loss L_com (few-shot) (ablation) | ACE (3 experts), CIFAR100-LT IF=100 · CIFAR100-LT IF=100 | few-shot accuracy: 17.5 -> 23.5 | Improved |
| use soft complementary loss instead of split-specific classifiers (SSC) (ablation) | ACE (3 experts), CIFAR100-LT IF=100 · CIFAR100-LT IF=100 | top-1 accuracy: 43.4 -> 49.4 | Improved |
| use linear learning-rate scaling instead of uniform learning rate across experts (ablation) | ACE (3 experts), CIFAR100-LT IF=100 · CIFAR100-LT IF=100 | top-1 accuracy: 41.7 -> 49.4 | Improved |
| use linear scaling instead of square-root scaling (ablation) | ACE (3 experts), CIFAR100-LT IF=100 · CIFAR100-LT IF=100 | top-1 accuracy: 49.1 -> 49.4 | No clear effect |
| use group average with scaling instead of group max (ablation) | ACE (3 experts), CIFAR100-LT IF=100 · CIFAR100-LT IF=100 | top-1 accuracy: 43.4 -> 49.4 | Improved |
| use group average with scaling instead of group concat (ablation) | ACE (3 experts), CIFAR100-LT IF=100 · CIFAR100-LT IF=100 | top-1 accuracy: 37.7 -> 49.4 | Improved |
| add logit scaling to group average aggregation (ablation) | ACE (3 experts), CIFAR100-LT IF=100 · CIFAR100-LT IF=100 | top-1 accuracy: 46.7 -> 49.4 | Improved |
| add logit scaling to group average aggregation (few-shot) (ablation) | ACE (3 experts), CIFAR100-LT IF=100 · CIFAR100-LT IF=100 | few-shot accuracy: 36.5 -> 23.5 | Hurt |
| use Mixup alpha 0.3 instead of no Mixup (alpha 0) (ablation) | ACE (3 experts), CIFAR100-LT IF=100 · CIFAR100-LT IF=100 | top-1 accuracy: 48.6 -> 49.4 | No clear effect |
| increase from 3 to 4 experts | ACE · CIFAR100-LT IF=100 | top-1 accuracy: 49.4 -> 49.6 | No clear effect |
Closest papers by meaning
- Deep Long-Tailed Learning: A Survey (TPAMI 2023 · similarity 0.89)
- Bag of Tricks for Long-Tailed Visual Recognition with Deep Convolutional Neural Networks (AAAI 2021 · similarity 0.89)
- Large-Scale Long-Tailed Recognition in an Open World (CVPR 2019 · similarity 0.89)
- U-MOF: Uncertainty-Guided Parameter-Efficient Multi-Objective Fine-Tuning for Long-Tailed Recognition (NeurIPS 2026 · similarity 0.89)
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.88)
- Decoupled Optimisation for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.87)
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.
