Research map / Label-efficient and robust learning / Recognition and long-tail
AutoMix: Mixup Networks for Sample Interpolation via Cooperative Barycenter Learning
Jianchao Zhu, Liangliang Shi, Junchi Yan, Hongyuan Zha
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 |
|---|---|---|---|
| use random-ratio label weighting (d) rather than first image's label (a) (ablation) | AutoMix · CIFAR-10 (full) | top-1 error: 5.38 -> 3.72 | Improved |
| use random-ratio label weighting rather than half-half mix (b) (ablation) | AutoMix · CIFAR-10 (full) | top-1 error: 3.91 -> 3.72 | Improved |
| use random-ratio label weighting rather than multi-label sum (c) (ablation) | AutoMix · CIFAR-10 (sub) | top-1 error: 15.35 -> 15.17 | Improved |
| mix images from random classes rather than only same-label pairs (ablation) | AutoMix · CIFAR-10 (full) | top-1 error: 4.21 -> 3.72 | Improved |
| mix images from random classes rather than only different-label pairs (ablation) | AutoMix · CIFAR-10 (full) | top-1 error: 3.98 -> 3.72 | Improved |
| add AutoMix over baseline | ResNet-18 · CIFAR-10 (full) | top-1 error: 4.59 -> 3.72 | Improved |
| add AutoMix over baseline | ResNet-18 · Tiny-ImageNet (full) | top-1 error: 40.35 -> 37.15 | Improved |
| add AutoMix over baseline | ResNet-18 · MIML | mAP: 72.14 -> 74.80 | Improved |
| add AutoMix over Manifold Mixup | ResNet-18 · GTSRB (full) | top-1 error: 0.38 -> 0.41 | No clear effect |
| add OptTransMix to BC | 11-layer CNN · MNIST (sub) | top-1 error: 5.93 -> 4.37 | Improved |
| add OptTransMix to Mixup | 11-layer CNN · F-MNIST (sub) | top-1 error: 17.38 -> 17.10 | Improved |
| add OptTransMix to Mixup | 11-layer CNN · MNIST (sub) | top-1 error: 6.17 -> 8.12 | Hurt |
Later papers that compared with this method
4 later accepted papers report AutoMix as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- TdAttenMix: Top-Down Attention Guided Mixup (AAAI 2025 · better on 7 of 7 shared results)
- Adversarial AutoMixup (ICLR 2024 · better on 17 of 17 shared results)
- SUMix: Mixup with Semantic and Uncertain Information (ECCV 2024 · better on 1 of 1 shared result)
- Harnessing Hard Mixed Samples with Decoupled Regularizer (NeurIPS 2023 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- AutoMix: Unveiling the Power of Mixup for Stronger Classifiers (ECCV 2022 · similarity 0.91)
- Adversarial AutoMixup (ICLR 2024 · similarity 0.88)
- A Unified Analysis of Mixed Sample Data Augmentation: A Loss Function Perspective (NeurIPS 2022 · similarity 0.87)
- AGMixup: Adaptive Graph Mixup for Semi-supervised Node Classification (AAAI 2025 · similarity 0.85)
- Supervision Interpolation via LossMix: Generalizing Mixup for Object Detection and Beyond (AAAI 2024 · similarity 0.85)
- Tailoring Mixup to Data for Calibration (ICLR 2025 · similarity 0.84)
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.
