Research map / Label-efficient and robust learning / Noisy labels
A Topological Filter for Learning with Label Noise
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 zeta-filtering (peeling the largest component) to the connected-component selection (ablation) | TopoCC · CIFAR-10, uniform 60% | test accuracy %: 78.7 ± 0.5 -> 80.5 ± 0.4 | Helped |
| Add zeta-filtering to TopoCC (ablation) | TopoCC · CIFAR-100, uniform 40% | test accuracy %: 57.3 ± 1.6 -> 62.0 ± 0.6 | Helped |
| Add zeta-filtering to TopoCC (ablation) | TopoCC · CIFAR-10, pair 40% | test accuracy %: 85.4 ± 0.4 -> 87.9 ± 0.2 | Helped |
| Replace Standard training with TopoFilter | ResNet-18 · CIFAR-10, uniform 60% | test accuracy %: 73.7 ± 1.1 -> 80.5 ± 0.4 | Helped |
| Replace Standard training with TopoFilter | ResNet-50 pretrained · Clothing1M | accuracy %: 68.94 -> 74.10 | Helped |
| Use noisy 10k validation set instead of clean 10k (ablation) | TopoFilter · CIFAR-10, uniform 60% | test accuracy %: 80.55 -> 80.18 | No clear effect |
| Use clean 1k validation set instead of clean 10k (ablation) | TopoFilter · CIFAR-10, uniform 60% | test accuracy %: 80.55 -> 80.16 | No clear effect |
| Vary k for largest connected component (kc 4 to 64) (ablation) | TopoFilter · CIFAR-10, uniform 60% | test accuracy %: 80.55 -> 79.68 (kc=32) | No clear effect |
| Vary k for outlier detection (ko 4 to 64) (ablation) | TopoFilter · CIFAR-10, uniform 60% | test accuracy %: 80.55 -> 79.93 (ko=4) | No clear effect |
| Vary feature dimension (32 to 512) (ablation) | TopoFilter · CIFAR-10, uniform 60% | test accuracy %: 80.55 -> 79.61 (dim 32) | No clear effect |
| Replace TopoFilter with Co-teaching at 80% uniform noise | CIFAR-10 · CIFAR-10, uniform 80% | test accuracy %: 22.9 ± 3.5 -> 45.7 ± 1.0 | Helped |
Closest papers by meaning
- Mitigating Label Noise on Graphs via Topological Sample Selection (ICML 2024 · similarity 0.88)
- FINE Samples for Learning with Noisy Labels (NeurIPS 2021 · similarity 0.88)
- On the Role of Label Noise in the Feature Learning Process (ICML 2025 · similarity 0.87)
- Directional Label Diffusion Model for Learning from Noisy Labels (CVPR 2025 · similarity 0.85)
- Topological Blindspots: Understanding and Extending Topological Deep Learning Through the Lens of Expressivity (ICLR 2025 · similarity 0.85)
- Handling Label Noise via Instance-Level Difficulty Modeling and Dynamic Optimization (NeurIPS 2025 · 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.