Image Classification on CIFAR-10 (Accuracy and Calibration)
96.74AccuracyDeep Ensembles
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Deep EnsemblesBackbone=WRN-28-10, Cutout=true2020.12 | 96.74 | 0.0093 | — | — | — | |
| MC dropoutBackbone=WRN-28-10, Cutout=true2020.12 | 96.5 | 0.0117 | — | — | — | |
| BatchEnsemblesBackbone=WRN-28-10, Cutout=true2020.12 | 96.48 | 0.0167 | — | — | — | |
| MCPBackbone=WRN-28-10, Cutout=true2020.12 | 96.33 | 0.0207 | — | — | — | |
| MC dropoutBackbone=WRN-28-10, Cutout=false2020.12 | 95.95 | 0.0172 | — | — | — | |
| LP-BNNBackbone=WRN-28-10, Cutout=true2020.12 | 95.02 | 0.0094 | — | — | — | |
| MIMOBackbone=WRN-28-10, Cutout=false2020.12 | 94.96 | 0.03 | — | — | — | |
| DUQBackbone=ResNet-18, Cutout=false2020.12 | 93.36 | 0.0131 | — | — | — | |
| DUQBackbone=WRN-28-10, Cutout=false2020.12 | 87.48 | 0.3983 | — | — | — | |
| EDLBackbone=WRN-28-10, Cutout=false2020.12 | 85.73 | 0.0904 | — | — | — | |
| ResNet20Model=ResNet202024.05 | 81.98 | 0.045 | 0.596 | 0.535 | 72.37 | |
| W-Asymmetric ResNet20Model=ResNet202024.05 | 81.94 | 0.044 | 0.6 | 0.535 | 73.64 | |
| W-Asymmetric ResNet110Model=ResNet1102024.05 | 77.4 | 0.049 | 0.745 | 0.658 | 63.2 | |
| ResNet110Model=ResNet1102024.05 | 75.71 | 0.052 | 0.803 | 0.706 | 59.85 | |
| W-Asymmetric MLP-8Model=MLP-82024.05 | 57.08 | 0.042 | 1.31 | 1.22 | 54.15 | |
| MLP-8Model=MLP-82024.05 | 56.37 | 0.039 | 1.34 | 1.24 | 52.87 | |
| W-Asymmetric MLP-16Model=MLP-162024.05 | 55.16 | 0.045 | 1.39 | 1.27 | 51.42 | |
| MLP-16Model=MLP-162024.05 | 13.54 | 0.026 | 2.29 | 2.28 | 13.34 |