Research map / Label-efficient and robust learning / Recognition and long-tail
An Easy-to-hard Learning Paradigm for Multiple Classes and Multiple Labels
Weiwei Liu, Ivor W. Tsang, Klaus-Robert Müller
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 easy-to-hard class chaining (CCMC) to OVR | OVR · KTH | testing error (%): 8.18 -> 7.34 | Improved |
| optimize class order with greedy search (ablation) | CCMC · KTH | testing error (%): 7.34 -> 6.29 | Improved |
| optimize class order with dynamic programming (ablation) | CCMC-Greedy · KTH | testing error (%): 6.29 -> 5.66 | Improved |
| add class chaining with DP ordering to OVR | OVR · IRIS | testing error (%): 10.00 -> 3.33 | Improved |
| add class chaining with DP ordering to OVR | OVR · RCV1 | testing error (%): 11.90 -> 11.68 | Improved |
| add class chaining with DP ordering to OVR with background data | OVR · SOWERBY | testing error (%): 18.03 -> 15.24 | Improved |
| increase fast-greedy parameter w from 10 to 100 (ablation) | CCMC-FG · ALOI | testing error (%): 5.27 -> 4.78 | Improved |
| increase fast-greedy parameter w from 10 to 100 (ablation) | CCMC-FG · ALOI | training time (s): 16,327s -> 163,891s | Hurt |
| use CCMC-DP instead of OVR | OVR · RCV1 | training time (s): 18.16s -> 16,450.55s | Hurt |
| add CCMC-FG on ResNet-34 features, w=50 | ResNet-34 · ILSVRC2012 subset | testing error (%): 21.51 -> 19.84 | Improved |
| add CCMC-FG on VGG features, w=10 | VGG · ILSVRC2012 subset | testing error (%): 23.95 -> 23.98 | No clear effect |
| increase w from 10 to 50 for CCMC-FG on VGG features (ablation) | CCMC-FG+VGG · ILSVRC2012 subset | testing error (%): 23.98 -> 20.73 | Improved |
Closest papers by meaning
- Self-Paced Unified Representation Learning for Hierarchical Multi-Label Classification (AAAI 2024 · similarity 0.85)
- Multi-class classification without multi-class labels (ICLR 2019 · similarity 0.85)
- Multi-Label Learning From Single Positive Labels (CVPR 2021 · similarity 0.85)
- Classifier-guided CLIP Distillation for Unsupervised Multi-label Classification (CVPR 2025 · similarity 0.83)
- Rethinking Self-Distillation: Label Averaging and Enhanced Soft Label Refinement with Partial Labels (ICLR 2025 · similarity 0.83)
- Multi-Label Open Set Recognition (NeurIPS 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.
