Research map / Label-efficient and robust learning / Recognition and long-tail
Improving Deep Regression with Ordinal Entropy
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 ordinal entropy L_d + L_t | NeW-CRFs · NYU-Depth-v2 | REL: 0.095 -> 0.089 | Helped |
| Add L_d + L_t | ResNet-50 depth baseline · NYU-Depth-v2 | δ1: 0.793 -> 0.811 | Helped |
| Add L_d + L_t | Crowd counting baseline · SHA | MAE: 68.2 -> 65.6 | Helped |
| Add L_d + L_t | Crowd counting baseline · SHB | MAE: 10.6 -> 9.1 | Helped |
| Add L_d + L_t | Age estimation baseline · AgeDB-DIR ALL | MAE: 7.77 -> 7.46 | Helped |
| Add unweighted diversity L'_d (ablation) | Regression baseline · Linear operator learning | MSE: 3.0 ± 0.93 -> 2.2 ± 0.42 | Helped |
| Use ordinally weighted L_d instead of unweighted L'_d (ablation) | Regression baseline · NYU-Depth-v2 | δ1: 0.795 -> 0.808 | Helped |
| Add the tightness term L_t (ablation) | Baseline + L_d · NYU-Depth-v2 | δ1: 0.808 -> 0.811 | Helped |
| Remove feature normalization (ablation) | Baseline + L_d · NYU-Depth-v2 | δ1: 0.808 -> 0.790 | Hurt |
| Replace L2 feature distance with cosine distance (ablation) | Baseline + L_d · NYU-Depth-v2 | δ1: 0.808 -> 0.806 | Hurt |
| Square the label-distance weights (ablation) | Baseline + L_d · NYU-Depth-v2 | δ1: 0.808 -> 0.800 | Hurt |
| Vary λ_t from 0 to 1 (ablation) | Age estimation with L_d · AgeDB-DIR ALL | MAE: 7.60 -> 7.46 | Helped |
Closest papers by meaning
- Improving Deep Regression with Tightness (ICLR 2025 · similarity 0.89)
- SLACE: A Monotone and Balance-Sensitive Loss Function for Ordinal Regression (AAAI 2025 · similarity 0.85)
- Generative Cross-Entropy: A Strictly Proper Loss for Data-Efficient Classification (NeurIPS 2026 · similarity 0.84)
- Neural Collapse in Cumulative Link Models for Ordinal Regression: An Analysis with Unconstrained Feature Model (NeurIPS 2025 · similarity 0.84)
- The Entropy Enigma: Success and Failure of Entropy Minimization (ICML 2024 · similarity 0.83)
- Engineering the Neural Collapse Geometry of Supervised-Contrastive Loss (Student Abstract) (AAAI 2024 · similarity 0.83)
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.