Confidence Calibration on CIFAR-100-LT (test)
0.015ECEKnowledge-Transferring-based Temperature Scaling
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Knowledge-Transferring-based Temperature ScalingModel=ResNet-322023.04 | 0.015 | — | — | — | — | — | — | — | |
| Knowledge-Transferring-based Temperature ScalingModel=VGG-192023.04 | 0.0199 | — | — | — | — | — | — | — | |
| ETSModel=ResNet-322023.04 | 0.021 | — | — | — | — | — | — | — | |
| ETSModel=VGG-192023.04 | 0.0236 | — | — | — | — | — | — | — | |
| Knowledge-Transferring-based Temperature ScalingModel=DenseNet-402023.04 | 0.0237 | — | — | — | — | — | — | — | |
| TSModel=ResNet-322023.04 | 0.025 | — | — | — | — | — | — | — | |
| ETSModel=DenseNet-402023.04 | 0.0251 | — | — | — | — | — | — | — | |
| GPCModel=DenseNet-402023.04 | 0.0273 | — | — | — | — | — | — | — | |
| GPCModel=ResNet-322023.04 | 0.0327 | — | — | — | — | — | — | — | |
| TSModel=DenseNet-402023.04 | 0.0343 | — | — | — | — | — | — | — | |
| TSModel=VGG-192023.04 | 0.0381 | — | — | — | — | — | — | — | |
| GPCModel=VGG-192023.04 | 0.0382 | — | — | — | — | — | — | — | |
| TS-IRModel=DenseNet-402023.04 | 0.0557 | — | — | — | — | — | — | — | |
| IROVAModel=DenseNet-402023.04 | 0.0576 | — | — | — | — | — | — | — | |
| IROVAModel=ResNet-322023.04 | 0.0592 | — | — | — | — | — | — | — | |
| SBCModel=DenseNet-402023.04 | 0.0596 | — | — | — | — | — | — | — | |
| TS-IRModel=ResNet-322023.04 | 0.0607 | — | — | — | — | — | — | — | |
| TS-IRModel=VGG-192023.04 | 0.0635 | — | — | — | — | — | — | — | |
| IROVAModel=VGG-192023.04 | 0.0666 | — | — | — | — | — | — | — | |
| SBCModel=ResNet-322023.04 | 0.0674 | — | — | — | — | — | — | — | |
| SBCModel=VGG-192023.04 | 0.0803 | — | — | — | — | — | — | — | |
| IRModel=DenseNet-402023.04 | 0.0842 | — | — | — | — | — | — | — | |
| IRModel=ResNet-322023.04 | 0.0935 | — | — | — | — | — | — | — | |
| IRModel=VGG-192023.04 | 0.1035 | — | — | — | — | — | — | — | |
| LADEimbalance ratio=502020.12 | 0.148 | 50.5 | 0.0043 | 0.658 | 2.02 | — | — | — | |
| Causal Normimbalance ratio=502020.12 | 0.15 | 48.1 | 0.0048 | 0.689 | 2.13 | — | — | — | |
| BaseModel=DenseNet-402023.04 | 0.16 | — | — | — | — | — | — | — | |
| Balanced Softmaximbalance ratio=502020.12 | 0.168 | 49.9 | 0.0046 | 0.673 | 2.07 | — | — | — | |
| PC Softmaximbalance ratio=502020.12 | 0.174 | 49.5 | 0.0047 | 0.678 | 2.1 | — | — | — | |
| BaseModel=ResNet-322023.04 | 0.2038 | — | — | — | — | — | — | — | |
| Softmaximbalance ratio=502020.12 | 0.249 | 45.5 | 0.0068 | 0.769 | 2.5 | — | — | — | |
| BaseModel=VGG-192023.04 | 0.2786 | — | — | — | — | — | — | — | |
| UniMix+BayiasBackbone=ResNet-322021.11 | 0.45 | — | — | 0.7824 | — | 0.517 | 0.623 | — | |
| BayiasBackbone=ResNet-322021.11 | 0.46 | — | — | 0.7878 | — | 0.518 | 0.631 | — | |
| UniMixBackbone=ResNet-322021.11 | 0.505 | — | — | 0.807 | — | 0.555 | 0.687 | — | |
| RemixBackbone=ResNet-322021.11 | 0.74 | — | — | 0.8933 | — | 0.675 | 0.846 | — | |
| mixupBackbone=ResNet-322021.11 | 0.751 | — | — | 0.8926 | — | 0.683 | 0.849 | — | |
| LDAM+DRWBackbone=ResNet-322021.11 | 0.801 | — | — | 1.074 | — | 0.671 | 1.093 | — | |
| ERMBackbone=ResNet-322021.11 | 0.831 | — | — | 0.9705 | — | 0.709 | 0.952 | — | |
| UniMix+BayiasBackbone=ResNet-32, Imbalance ratio (rho)=102021.11 | 15.24 | — | — | — | — | — | — | 23.67 | |
| BayiasBackbone=ResNet-32, Imbalance ratio (rho)=102021.11 | 17.42 | — | — | — | — | — | — | 28.26 | |
| UniMixBackbone=ResNet-32, Imbalance ratio (rho)=102021.11 | 19.38 | — | — | — | — | — | — | 33.4 | |
| mixupBackbone=ResNet-32, Imbalance ratio (rho)=102021.11 | 19.69 | — | — | — | — | — | — | 38.53 | |
| RemixBackbone=ResNet-32, Imbalance ratio (rho)=102021.11 | 20.17 | — | — | — | — | — | — | 32.99 | |
| ERMBackbone=ResNet-32, Imbalance ratio (rho)=102021.11 | 22.85 | — | — | — | — | — | — | 34.5 | |
| UniMix+BayiasBackbone=ResNet-32, Imbalance ratio (rho)=1002021.11 | 23.04 | — | — | — | — | — | — | 37.36 | |
| BayiasBackbone=ResNet-32, Imbalance ratio (rho)=1002021.11 | 24.31 | — | — | — | — | — | — | 39.66 | |
| UniMixBackbone=ResNet-32, Imbalance ratio (rho)=1002021.11 | 27.12 | — | — | — | — | — | — | 41.46 | |
| LDAM+DRWBackbone=ResNet-32, Imbalance ratio (rho)=102021.11 | 30.54 | — | — | — | — | — | — | 55.57 | |
| mixupBackbone=ResNet-32, Imbalance ratio (rho)=1002021.11 | 32.72 | — | — | — | — | — | — | 50.46 | |
| RemixBackbone=ResNet-32, Imbalance ratio (rho)=1002021.11 | 33.56 | — | — | — | — | — | — | 50.96 | |
| ERMBackbone=ResNet-32, Imbalance ratio (rho)=1002021.11 | 38.23 | — | — | — | — | — | — | 87.22 | |
| LDAM+DRWBackbone=ResNet-32, Imbalance ratio (rho)=1002021.11 | 42.18 | — | — | — | — | — | — | 64.78 |