Research map / Label-efficient and robust learning / Recognition and long-tail
Dual Perspective of Label-Specific Feature Learning for Multi-Label 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 |
|---|---|---|---|
| label-specific non-informative feature identification and perturbation (DELA) vs. shared-representation classifiers without noise (DELA-nn) (ablation) | DELA · 10 benchmark multi-label data sets | Average precision: DELA-nn lower on most datasets, e.g. rcv1-s1 0.6114 vs 0.6391 | Helped |
| identifying label-specific non-informative features (vs. shared noise only, DELA-sn) (ablation) | DELA · 10 benchmark multi-label data sets | Average precision: DELA-sn 0.3338 vs DELA 0.3382 on corel5k; win overall p=0.0020 | Helped |
| label-specific feature selection on shared representation (DELA-fs) instead of perturbation-based immutability (ablation) | DELA · 10 benchmark multi-label data sets | Average precision: DELA-fs 0.3313 vs DELA 0.3382 on corel5k; win overall p=0.0039 | Helped |
| DELA vs. deep label-specific feature transformation PACA | DELA · 10 benchmark data sets, six metrics | Average precision / Macro AUC / Hamming loss: Macro AUC tie with PACA (p=0.0059 tie), win on Average precision and Hamming loss | Mixed |
| DELA vs. LIFT, LLSF, C2AE, MPVAE, CLIF baselines | DELA · 10 benchmark data sets | Average precision: DELA wins on all 10 datasets for Average precision (Wilcoxon win vs all baselines) | Helped |
| trade-off parameter beta set to 0 (removing the KL noise-level constraint) (ablation) | DELA · validation, varying beta | Average precision: degraded performance when beta = 0 (no numeric values reported) | Hurt |
Closest papers by meaning
- End-to-End Probabilistic Label-Specific Feature Learning for Multi-Label Classification (AAAI 2022 · similarity 0.93)
- Double-Layer Hybrid-Label Identification Feature Selection for Multi-View Multi-Label Learning (AAAI 2024 · similarity 0.88)
- Generalization Analysis for Label-Specific Representation Learning (NeurIPS 2024 · similarity 0.87)
- Imprecise Label Learning: A Unified Framework for Learning with Various Imprecise Label Configurations (NeurIPS 2024 · similarity 0.87)
- Compositional Targeted Multi-Label Universal Perturbations (CVPR 2025 · similarity 0.87)
- Addressing Multi-Label Learning with Partial Labels: From Sample Selection to Label Selection (AAAI 2025 · similarity 0.86)
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.