Image Classification on CIFAR-100 (Symmetric/Asymmetric Test Accuracy)
77.9Test Acc (symm 20%)SAT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SATBackbone=ResNet34, Learning rate scheduler=cosine annealing, Trials=32021.08 | 77.9 | 75.38 | — | — | 73.73 | |
| CARBackbone=ResNet34, Learning rate scheduler=cosine annealing, Trials=32021.08 | 77.9 | 75.38 | 69.78 | 38.24 | 74.89 | |
| ELRBackbone=ResNet34, Learning rate scheduler=cosine annealing, Trials=32021.08 | 74.68 | 68.43 | 60.05 | 30.27 | 73.73 | |
| NLNLBackbone=ResNet34, Trials=32021.08 | 73.55 | 66.92 | 56.51 | 32.16 | 62.69 | |
| SELFBackbone=ResNet34, Trials=32021.08 | 73.55 | 66.92 | 56.51 | 32.16 | 62.69 | |
| SLBackbone=ResNet34, Trials=32021.08 | 71.52 | 66.39 | 57.17 | 35.56 | 38.72 | |
| DACBackbone=ResNet34, Trials=32021.08 | 70.38 | 66.71 | 53.16 | 29.16 | 47.22 | |
| BootstrapBackbone=ResNet34, Trials=32021.08 | 66.81 | 60.11 | 47.94 | 17.18 | 56.51 | |
| GCEBackbone=ResNet34, Trials=32021.08 | 61.77 | 62.27 | 34.68 | 21.64 | 45.12 | |
| Cross EntropyBackbone=ResNet34, Trials=32021.08 | 58.72 | 48.2 | 39.19 | 19.12 | 42.74 | |
| Forward TBackbone=ResNet34, Trials=32021.08 | 58.27 | 47.66 | 37.41 | 18.1 | 34.44 | |
| Joint OptBackbone=ResNet34, Trials=32021.08 | 58.15 | 54.81 | 45.7 | 25.91 | 69.32 |