Research map / Label-efficient and robust learning / Recognition and long-tail
Unimodal-Concentrated Loss: Fully Adaptive Label Distribution Learning for Ordinal Regression
Check 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 unimodal loss instead of softmax as the auxiliary loss to concentrated loss (ablation) | Unimodal-concentrated loss · MORPH II, AFLW2000, and BIWI | MAE: 1.92 -> 1.86; 4.25 -> 4.13; 3.61 -> 3.57 | Helped |
| Use concentrated loss instead of mean and variance losses, with unimodal auxiliary loss (ablation) | Ordinal regression model · MORPH II, AFLW2000, and BIWI | MAE: 3.30 -> 1.86; 4.53 -> 4.13; 4.39 -> 3.57 | Helped |
| Add concentrated loss to unimodal loss (ablation) | Ordinal regression model · MORPH II, AFLW2000, and BIWI | MAE: ≥ 5 -> 1.86; ≥ 7 -> 4.13; ≥ 7 -> 3.57 | Helped |
| Add an auxiliary loss to concentrated loss alone (ablation) | Ordinal regression model · MORPH II, AFLW2000, and BIWI | MAE: NaN -> 1.86; NaN -> 4.13; NaN -> 3.57 | Helped |
| Replace DLDL-v2 with unimodal-concentrated loss | VGG-16 ordinal regression model · MORPH II, AFLW2000, and BIWI | MAE: 1.90 -> 1.86; 4.20 -> 4.13; 3.80 -> 3.57 | Helped |
| Replace Mean-Variance with unimodal-concentrated loss | VGG-16 ordinal regression model · MORPH II, AFLW2000, and BIWI | MAE: 2.01 -> 1.86; 4.36 -> 4.13; 4.01 -> 3.57 | Helped |
| Set λ to 1e-1 instead of the reported operating value 1e3 (ablation) | Unimodal-concentrated loss · MORPH II, AFLW2000, and BIWI | MAE: 1.86 -> NaN; 4.13 -> NaN; 3.57 -> NaN | Hurt |
| Set λ to 1e4 instead of the reported operating value 1e3 (ablation) | Unimodal-concentrated loss · MORPH II, AFLW2000, and BIWI | MAE: 1.86 -> 3.24; 4.13 -> 6.12; 3.57 -> 5.75 | Hurt |
| Set λ to 1e2 rather than 1e3 (ablation) | Unimodal-concentrated loss · MORPH II | MAE: 1.86 -> 1.88 | Hurt |
| Set λ to 2e3 rather than 1e3 (ablation) | Unimodal-concentrated loss · AFLW2000 | MAE: 4.13 -> 4.11 | Helped |
Closest papers by meaning
- SLACE: A Monotone and Balance-Sensitive Loss Function for Ordinal Regression (AAAI 2025 · similarity 0.86)
- Label Shift Meets Online Learning: Ensuring Consistent Adaptation with Universal Dynamic Regret (CVPR 2025 · similarity 0.85)
- L2B: Learning to Bootstrap Robust Models for Combating Label Noise (CVPR 2024 · similarity 0.84)
- Forming Auxiliary High-confident Instance-level Loss to Promote Learning from Label Proportions (CVPR 2025 · similarity 0.84)
- Learning from Noisy Labels via Conditional Distributionally Robust Optimization (NeurIPS 2024 · similarity 0.84)
- L-WISE: Boosting Human Visual Category Learning Through Model-Based Image Selection and Enhancement (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.