Image Classification on CIFAR-100 20% symmetric label noise
81.6Peak AccuracyC2D (DivideMix with SimCLR)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| C2D (DivideMix with SimCLR)Architecture=ResNet-502021.03 | 81.6 | 80.89 | |
| C2D (ELR+ with SimCLR)Architecture=ResNet-342021.03 | 79.18 | 79.03 | |
| ODDArchitecture=WRN-28-102021.03 | 79.1 | — | |
| C2D (DivideMix with SimCLR)Architecture=PreAct ResNet-182021.03 | 78.69 | 78.32 | |
| DivideMixArchitecture=PreAct ResNet-182021.03 | 77.3 | 76.9 | |
| CE+mixup with SimCLRArchitecture=ResNet-342021.03 | 76.46 | 76.34 | |
| Meta-learningArchitecture=PreAct ResNet-322021.03 | 68.5 | 67.7 | |
| ELR+Architecture=ResNet-342021.03 | — | 77.6 |