Research map / Label-efficient and robust learning / Noisy labels
Switchable K-Class Hyperplanes for Noise-Robust Representation Learning
Boxiao Liu, Guanglu Song, Manyuan Zhang, Haihang You, Yu Liu
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 SKH (switchable hyperplanes, M=3) to ArcFace | ArcFace · IJB-C, MS1MV0 (noisy) | TAR@FAR=1e-4: 91.82 -> 95.25 | Improved |
| add SKH (switchable hyperplanes, M=3) to ArcFace | ArcFace · IJB-B, MS1MV0 (noisy) | TAR@FAR=1e-4: 89.24 -> 93.50 | Improved |
| use SKH instead of co-mining | ArcFace · IJB-C, MS1MV0 (noisy) | TAR@FAR=1e-4: 93.82 -> 95.25 | Improved |
| use SKH instead of sub-center ArcFace | ArcFace · IJB-C, MS1MV0 (noisy) | TAR@FAR=1e-4: 93.72 -> 95.25 | Improved |
| increase number of hyperplanes M from 3 to 5 (ablation) | SKH + ArcFace · IJB-C, MS1MV0 | TAR@FAR=1e-4: 95.25 -> 95.04 | Hurt |
| reduce number of hyperplanes M from 3 to 2 (ablation) | SKH + ArcFace · IJB-C, MS1MV0 | TAR@FAR=1e-4: 95.25 -> 94.63 | Hurt |
| add posterior data clean strategy (ablation) | SKH + ArcFace M=3 · IJB-B, MS1MV0 | TAR@FAR=1e-4: 93.50 -> 94.82 | Improved |
| add posterior data clean strategy (ablation) | SKH + ArcFace M=3 · IJB-C, MS1MV0 | TAR@FAR=1e-4: 95.25 -> 96.26 | Improved |
| vary thresholds T1 and T2 of posterior clean (ablation) | SKH M=3 · IJB-C, MS1MV0 | TAR@FAR=1e-4: 96.12 -> 96.26 | No clear effect |
| add SKH to ArcFace under inter-class noise | ArcFace · IJB-B, IT-MS1MV3 | TAR@FAR=1e-4: 86.96 -> 94.81 | Improved |
| add SKH to ArcFace under intra-class noise | ArcFace · IJB-C, IA-MS1MV3 | TAR@FAR=1e-4: 95.31 -> 96.06 | Improved |
| add SKH to ArcFace on clean MS1MV3 | ArcFace · IJB-C, MS1MV3 (clean) | TAR@FAR=1e-4: 96.44 -> 95.25 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- ArcFace — ArcFace: Additive Angular Margin Loss for Deep Face Recognition (CVPR 2019) · better on 4 of 4 shared results
- CosFace — CosFace: Large Margin Cosine Loss for Deep Face Recognition (CVPR 2018) · 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
- Hinge-Minimax Learner for the Ensemble of Hyperplanes (JMLR 2018 · similarity 0.85)
- Learning with Open-world Noisy Data via Class-independent Margin in Dual Representation Space (AAAI 2025 · similarity 0.84)
- I-Con: A Unifying Framework for Representation Learning (ICLR 2025 · similarity 0.83)
- Trading off Consistency and Dimensionality of Convex Surrogates for Multiclass Classification (NeurIPS 2024 · similarity 0.83)
- Ditch the Denoiser: Emergence of Noise Robustness in Self-Supervised Learning from Data Curriculum (NeurIPS 2025 · similarity 0.82)
- Leveraging Data Symmetries to Select an Optimal Subset of Training Data under Label Noise (NeurIPS 2026 · similarity 0.82)
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.
