Research map / Label-efficient and robust learning / Recognition and long-tail
Enhancing Minority Classes by Mixing: An Adaptative Optimal Transport Approach for Long-tailed Classification
Jintong Gao, He Zhao, Zhuo Li, Dan dan Guo
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 OTmix to ERM | ERM · CIFAR-LT-100 imb100 | top-1 error (%): 63.2 -> 53.6 | Improved |
| replace random CMO pairing with OT-based pairing | ERM + CMO · CIFAR-LT-10 imb100 | top-1 error (%): 25.0 -> 21.7 | Improved |
| add OTmix to RIDE (3 experts) | RIDE · ImageNet-LT | top-1 error (%): 45.1 -> 42.7 | Improved |
| add OTmix to ERM | ERM · iNaturalist 2018 | top-1 error (%): 39.0 -> 30.5 | Improved |
| add OTmix to Cutmix | Cutmix · CIFAR-100 balanced | top-1 error (%): 31.6 -> 28.0 | Improved |
| add OTmix to SaliencyMix | SaliencyMix · CIFAR-100 balanced | top-1 error (%): 31.2 -> 21.7 | Improved |
| use sample-level cost only (ablation) | ERM + OTmix · CIFAR-LT-100 imb100 | top-1 error (%): 53.6 -> 55.1 | Hurt |
| use class-level cost only (ablation) | ERM + OTmix · CIFAR-LT-100 imb100 | top-1 error (%): 53.6 -> 54.5 | Hurt |
| euclidean distance instead of cosine similarity (ablation) | ERM + OTmix · CIFAR-LT-100 imb100 | top-1 error (%): 53.6 -> 54.1 | Hurt |
| keep same-category pairs (do not zero confusion-matrix diagonal) (ablation) | ERM + OTmix · CIFAR-LT-100 imb100 | top-1 error (%): 53.6 -> 54.9 | Hurt |
| sample background by probability instead of max transport value (ablation) | ERM + OTmix · CIFAR-LT-100 imb100 | top-1 error (%): 53.6 -> 55.0 | Hurt |
| pair by confusion matrix instead of OT (ablation) | ERM + OTmix · CIFAR-LT-100 imb100 | top-1 error (%): 53.6 -> 57.1 | Hurt |
Closest papers by meaning
- The Majority Can Help the Minority: Context-Rich Minority Oversampling for Long-Tailed Classification (CVPR 2022 · similarity 0.91)
- Enhance Image Classification via Inter-Class Image Mixup with Diffusion Model (CVPR 2024 · similarity 0.86)
- BEM: Balanced and Entropy-based Mix for Long-Tailed Semi-Supervised Learning (CVPR 2024 · similarity 0.86)
- Decoupled Optimisation for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.85)
- Catch-Up Mix: Catch-Up Class for Struggling Filters in CNN (AAAI 2024 · similarity 0.85)
- Long-Tailed Learning as Multi-Objective Optimization (AAAI 2024 · 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.
