Research map / Label-efficient and robust learning / Recognition and long-tail
Top-k Hierarchical Classification
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 |
|---|---|---|---|
| assign each node's reward to a single child and account for joint selection (K-H instead of H-only top-k rewards) (ablation) | hierarchical top-k selection · Gene function cellcyle, K=5 | misclassification cost: 79.11 -> 52.47 | Helped |
| assign each node's reward to a single child and account for joint selection (K-H instead of H-only top-k rewards) (ablation) | hierarchical top-k selection · Expertise matching, K=5, c2=0 | misclassification cost: 20.54 -> 9.67 | Helped |
| use hierarchical reward (K-H) instead of plain probability ranking (K-only) (ablation) | top-k selection · Gene function cellcyle, K=10 | misclassification cost: 48.19 -> 39.43 | Helped |
| use hierarchical reward (K-H) instead of plain probability ranking (K-only) at K=1 (ablation) | single selection · Gene function (12 datasets), K=1 | misclassification cost: 90.97 -> 90.01 (cellcyle) | Mixed |
| use hierarchical reward (K-H) instead of plain probability ranking (K-only) (ablation) | top-k selection · Expertise matching, K=3, c2=60 | misclassification cost: 23.20 -> 22.04 | Mixed |
| set partial-match cost c2=100 (equal to full mismatch) (ablation) | K-H vs baselines · Expertise matching, K=5 | misclassification cost: 19.10 for all three | No clear effect |
| use K-H instead of K-only under the standard top-k (full match) criterion | top-k selection · Expertise matching, K=5, c2=0 full match % | full match %: 80.90 -> 76.83 | Hurt |
Closest papers by meaning
- Hierarchical classification at multiple operating points (NeurIPS 2022 · similarity 0.90)
- Learning with Average Top-k Loss (NeurIPS 2017 · similarity 0.88)
- To Each Metric Its Decoding: Post-Hoc Optimal Decision Rules of Probabilistic Hierarchical Classifiers (ICML 2025 · similarity 0.86)
- Cardinality-Aware Set Prediction and Top-$k$ Classification (NeurIPS 2024 · similarity 0.84)
- Improved Balanced Classification with Theoretically Grounded Loss Functions (NeurIPS 2025 · similarity 0.83)
- Average Sensitivity of Hierarchical $k$-Median Clustering (ICML 2025 · 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.