Research map / Label-efficient and robust learning / Recognition and long-tail
Hierarchical Multi-Label Classification Networks
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 local outputs (per-level losses) to global-only deep network (ablation) | HMCN-F · FunCat validation (8 datasets) | average ranking AU(PRC): average ranking 5.25 -> 3.00 | Helped |
| add input feature reuse at each layer (ablation) | HMCN-F · FunCat validation (8 datasets) | average ranking AU(PRC): average ranking 5.25 -> 3.38 | Helped |
| combine input reuse and local outputs (ablation) | HMCN-F · FunCat validation (8 datasets) | average ranking AU(PRC): average ranking 1.25 (best of four variants) | Helped |
| hierarchical violation penalty lambda=0.1 added (ablation) | HMCN-F · FunCat validation (8 datasets) | AU(PRC): slightly better, e.g. CELLCYCLE 0.250 -> 0.252 (lambda=1.0 hurts: 0.233) | Helped |
| hierarchical violation penalty lambda=0.1 added (ablation) | HMCN-R · FunCat validation (8 datasets) | AU(PRC): e.g. CELLCYCLE 0.245 -> 0.249; GASCH2 0.244 -> 0.249 | Helped |
| increase hierarchical violation penalty to lambda=1.0 (ablation) | HMCN-F · FunCat validation (8 datasets) | AU(PRC): e.g. EXPR 0.301 (lambda=0.1) -> 0.263 | Hurt |
| replace baselines with HMCN-F (local+global, violation penalty) on FunCat test | HMCN-F · FunCat test | AU(PRC): CELLCYCLE 0.252 vs best baseline CSSA 0.188 / Clus-Ens 0.227 | Helped |
| HMCN-F with local+global optimization vs Clus-Ens on Reuters | HMCN-F · Reuters test | AU(PRC): 0.649 vs Clus-Ens 0.703 | Hurt |
| use recurrent LSTM-style HMCN-R instead of feed-forward HMCN-F | HMCN-R · FunCat/GO/other test (21 datasets) | average ranking AU(PRC): average ranking 2.04 vs HMCN-F 1.07; HMCN-R ~3M params vs HMCN-F ~7M on Cellcycle GO | Mixed |
Closest papers by meaning
- Coherent Hierarchical Multi-Label Classification Networks (NeurIPS 2020 · similarity 0.92)
- Self-Paced Unified Representation Learning for Hierarchical Multi-Label Classification (AAAI 2024 · similarity 0.90)
- Label Relation Graphs Enhanced Hierarchical Residual Network for Hierarchical Multi-Granularity Classification (CVPR 2022 · similarity 0.88)
- MLC-NC: Long-Tailed Multi-Label Image Classification Through the Lens of Neural Collapse (AAAI 2025 · similarity 0.85)
- Taming the Sigmoid Bottleneck: Provably Argmaxable Sparse Multi-Label Classification (AAAI 2024 · similarity 0.84)
- Enhancing Multi-Label Classification via Dynamic Label-Order Learning (AAAI 2024 · 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.