Research map / Label-efficient and robust learning / Recognition and long-tail
Confident Multiple Choice Learning
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 |
|---|---|---|---|
| replace MCL's oracle-only objective with confident oracle loss | 5 simple CNNs · CIFAR-10 test | top-1 error rate: 60.40% -> 15.65% | Helped |
| add feature sharing to confident-oracle-loss CMCL (ablation) | 5 simple CNNs · CIFAR-10 test | top-1 error rate: 15.65% -> 14.83% | Helped |
| add stochastic labeling to CMCL with feature sharing (ablation) | 5 simple CNNs · CIFAR-10 test | top-1 error rate: 14.83% -> 14.78% | Helped |
| use CMCL instead of IE | 5 ResNet-20 models · CIFAR-10 test | top-1 error rate: 10.18% -> 8.75% (-14.05%) | Helped |
| use CMCL instead of IE | 5 ResNet-20 models · SVHN test | top-1 error rate: 3.94% -> 3.68% (-6.60%) | Helped |
| use CMCL instead of IE | 5 VGGNet-17 models · CIFAR-10 test | top-1 error rate: 8.21% -> 7.83% (-4.63%) | Helped |
| use CMCL instead of IE | 5 VGGNet-17 models · SVHN test | top-1 error rate: 4.10% -> 3.92% (-4.39%) | Helped |
| use CMCL instead of IE | 5 GoogLeNet-18 models · CIFAR-10 test | top-1 error rate: 7.97% -> 7.51% (-5.77%) | Helped |
| use CMCL instead of IE | 5 GoogLeNet-18 models · SVHN test | top-1 error rate: 3.60% -> 3.44% (-4.44%) | Helped |
| increase overlap parameter K from 1 to 2 (ablation) | CMCL, M=5 · CIFAR-10 test | top-1 error rate: 14.78% -> 14.25% (-7.11%) | Helped |
| increase overlap parameter K from 2 to 3 (ablation) | CMCL, M=5 · CIFAR-10 test | top-1 error rate: 14.25% -> 14.38% | Hurt |
| increase overlap parameter K from 3 to 4 (ablation) | CMCL, M=5 · CIFAR-10 test | top-1 error rate: 14.38% -> 14.49% | Hurt |
Closest papers by meaning
- Versatile Multiple Choice Learning and Its Application to Vision Computing (CVPR 2019 · similarity 0.90)
- Confidence-aware Contrastive Learning for Selective Classification (ICML 2024 · similarity 0.83)
- Batch Selection for Multi-Label Classification Guided by Uncertainty and Dynamic Label Correlations (AAAI 2025 · similarity 0.82)
- Bayesian Low-Rank Learning (Bella): A Practical Approach to Bayesian Neural Networks (AAAI 2025 · similarity 0.82)
- Mitigating Label Bias in Machine Learning: Fairness through Confident Learning (AAAI 2024 · similarity 0.81)
- Improving Accuracy and Calibration via Differentiated Deep Mutual Learning (CVPR 2025 · similarity 0.81)
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.