Research map / Label-efficient and robust learning / Recognition and long-tail
Multi-class classification without multi-class labels
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 |
|---|---|---|---|
| Use MCL with pairwise similarity labels instead of class-label cross-entropy | ResNet101 · CIFAR10 test | error rate: 6.5% -> 5.6% | Helped |
| Use MCL with pairwise similarity labels instead of class-label cross-entropy | VGG8 · CIFAR100 test | error rate: 35.4% -> 36.1% | Hurt |
| Use MCL instead of KCL | ResNet101 · CIFAR10 test | error rate: 79.9% -> 5.6% | Helped |
| Use MCL instead of KCL | LeNet · MNIST test | error rate: 0.5% -> 0.6% | Hurt |
| Use MCL instead of KCL | cross-task clustering · Omniglot eval (K=C) | ACC: 82.4% -> 83.3% | Helped |
| Use MCL instead of KCL when the class count is unknown | cross-task clustering · Omniglot eval (K=100) | ACC: 78.1% -> 80.2% | Helped |
| Use MCL instead of KCL | cross-task clustering · ImageNet118 target (K=30) | ACC: 73.8% -> 74.4% | Helped |
| Use MCL instead of KCL when the class count is unknown | cross-task clustering · ImageNet118 target (K=100) | ACC: 65.2% -> 71.5% | Helped |
| Optimize the similarity predictor jointly with the classifier rather than use a static predictor | semi-supervised MCL · CIFAR10 4k labels | test error rate: 22.8% -> 18.0% | Helped |
| Use Pseudo-MCL rather than VAT | ResNet-18 · CIFAR10 4k labels | test error rate: 18.2 $\pm$ 0.4\% -> 18.0 $\pm$ 0.4\% | Helped |
Closest papers by meaning
- Learning from Complementary Labels (NeurIPS 2017 · similarity 0.88)
- Consistent Multi-Class Classification from Multiple Unlabeled Datasets (ICLR 2024 · similarity 0.88)
- Reliable Multilabel Classification: Prediction with Partial Abstention (AAAI 2020 · similarity 0.86)
- Learning to Learn with Conditional Class Dependencies (ICLR 2019 · similarity 0.86)
- Theoretical Insights Into Multiclass Classification: A High-dimensional Asymptotic View (NeurIPS 2020 · similarity 0.86)
- Taming the Sigmoid Bottleneck: Provably Argmaxable Sparse Multi-Label Classification (AAAI 2024 · similarity 0.85)
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.