Research map / Label-efficient and robust learning / Noisy labels
When Optimizing $f$-Divergence is Robust 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 |
|---|---|---|---|
| Optimize Total Variation divergence instead of the compared baselines | Noisy-label classification · CIFAR-10, Uniform high noise | accuracy | Helped |
| Optimize an f-divergence on noisy labels rather than use the strongest baseline | Noisy-label classification · CIFAR-10, Sparse high noise | accuracy | Hurt |
| Optimize an f-divergence on noisy labels rather than use the strongest baseline | Noisy-label classification · CIFAR-100, Sparse noise | accuracy | Hurt |
| Use bias correction for Jensen-Shannon divergence (ablation) | Jensen-Shannon on CIFAR-10 · Uniform low noise | accuracy: +0.64 | Helped |
| Use bias correction for Pearson $\mathcal X^2$ divergence (ablation) | Pearson $\mathcal X^2$ on CIFAR-10 · Uniform high noise | accuracy: +0.18 | Helped |
| Use bias correction for Pearson $\mathcal X^2$ divergence (ablation) | Pearson $\mathcal X^2$ on CIFAR-10 · Uniform low noise | accuracy: -0.12 | Hurt |
| Use bias correction for Jensen-Shannon divergence (ablation) | Jensen-Shannon on CIFAR-10 · Uniform high noise | accuracy: -0.18 | Hurt |
| Use bias correction for KL divergence (ablation) | KL on CIFAR-10 · Uniform low noise | accuracy: -1.34 | Hurt |
| Use bias correction for KL divergence (ablation) | KL on CIFAR-10 · Uniform high noise | accuracy: -0.38 | Hurt |
| Optimize Total Variation divergence rather than CE on clean data | Clean-data classification · MNIST | accuracy: -0.04 | No clear effect |
| Optimize Total Variation divergence rather than CE on clean data | Clean-data classification · CIFAR-10 | accuracy: -0.81 | Hurt |
| Use Total Variation divergence for noisy-label training | Classifier over arbitrary hypothesis space · Label-noise classification | robustness of the maximizer | Helped |
Closest papers by meaning
- $f$-Divergence Based Classification: Beyond the Use of Cross-Entropy (ICML 2024 · similarity 0.87)
- Generalized Jensen-Shannon Divergence Loss for Learning with Noisy Labels (NeurIPS 2021 · similarity 0.87)
- Variation-Bounded Loss for Noise-Tolerant Learning (AAAI 2026 · similarity 0.87)
- Label Noise: Ignorance Is Bliss (NeurIPS 2024 · similarity 0.86)
- Robust Diffusion Models via Divergence-Induced Weighted Denoising (NeurIPS 2026 · similarity 0.85)
- $\epsilon$-Softmax: Approximating One-Hot Vectors for Mitigating Label Noise (NeurIPS 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.