Research map / Label-efficient and robust learning / Noisy labels
SplitNet: Learnable Clean-Noisy Label Splitting for Learning with Noisy Labels
Daehwan Kim, Kwangrok Ryoo, Hansang Cho, Seungryong Kim
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 SplitNet (ablation) | SplitNet framework · CIFAR-100 90% noise | accuracy: 33.4 -> 50.7 | Improved |
| add SplitNet (ablation) | SplitNet framework · CIFAR-100 80% noise | accuracy: 59.1 -> 70.3 | Improved |
| add SplitNet (ablation) | SplitNet framework · CIFAR-100 20% noise | accuracy: 76.8 -> 80.6 | Improved |
| add proposed warm-up (K-fold cross-filtering and SSL boosting) (ablation) | SplitNet framework · CIFAR-100 90% noise | accuracy: 41.0 -> 50.7 | Improved |
| add proposed warm-up (ablation) | SplitNet framework · CIFAR-100 20% noise | accuracy: 79.8 -> 80.6 | Improved |
| add risk hedging (ablation) | SplitNet framework · CIFAR-100 90% noise | accuracy: 46.1 -> 50.7 | Improved |
| add risk hedging (ablation) | SplitNet framework · CIFAR-100 50% noise | accuracy: 77.5 -> 77.8 | Improved |
| use SplitNet framework instead of DivideMix with FixMatch (ablation) | SplitNet · CIFAR-100 90% noise | accuracy: 26.5 -> 50.7 | Improved |
| use dynamic threshold by split confidence instead of fixed 0.5 or 0.95 (ablation) | SplitNet SSL · CIFAR-100 90% noise | pseudo-label correctness | Improved |
| include prediction difference as SplitNet input (ablation) | SplitNet · CIFAR-100 | accuracy | Improved |
| use batch normalization in SplitNet (ablation) | SplitNet · CIFAR-100 | convergence | Improved |
| use 3 layers instead of 2 or 4 in SplitNet (ablation) | SplitNet · CIFAR-100 | accuracy | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DivideMix — DivideMix: Learning with Noisy Labels as Semi-supervised Learning (ICLR 2020) · better on 11 of 11 shared results
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
- CrossSplit: Mitigating Label Noise Memorization through Data Splitting (ICML 2023 · similarity 0.90)
- DivideMix: Learning with Noisy Labels as Semi-supervised Learning (ICLR 2020 · similarity 0.89)
- BadLabel: A Robust Perspective on Evaluating and Enhancing Label-Noise Learning (TPAMI 2024 · similarity 0.89)
- Learning with Structural Labels for Learning with Noisy Labels (CVPR 2024 · similarity 0.89)
- Understanding and Utilizing Deep Neural Networks Trained with Noisy Labels (ICML 2019 · similarity 0.88)
- Toward Robustness against Label Noise in Training Deep Discriminative Neural Networks (NeurIPS 2017 · similarity 0.88)
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.
