Image Classification on CIFAR-10 (test) (Accuracy and Weights Remaining)
96.1AccuracyDivideMix
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DivideMixNoise type=Symmetric, Noise ratio=20%2024.08 | 96.1 | — | |
| OursNoise type=Symmetric, Noise ratio=20%2024.08 | 95.92 | — | |
| ELR+Noise type=Symmetric, Noise ratio=20%2024.08 | 95.8 | — | |
| OursNoise type=Symmetric, Noise ratio=50%2024.08 | 95.67 | — | |
| SelCL+Noise type=Symmetric, Noise ratio=20%2024.08 | 95.5 | — | |
| OursNoise type=Symmetric, Noise ratio=80%2024.08 | 95.04 | — | |
| TCLNoise type=Symmetric, Noise ratio=20%2024.08 | 95 | — | |
| OursNoise type=Assymetric, Noise ratio=40%2024.08 | 94.89 | — | |
| ELR+Noise type=Symmetric, Noise ratio=50%2024.08 | 94.8 | — | |
| DivideMixNoise type=Symmetric, Noise ratio=50%2024.08 | 94.6 | — | |
| OursNoise type=Symmetric, Noise ratio=90%2024.08 | 94.23 | — | |
| LossModellingNoise type=Symmetric, Noise ratio=20%2024.08 | 94 | — | |
| SelCL+Noise type=Symmetric, Noise ratio=50%2024.08 | 93.9 | — | |
| TCLNoise type=Symmetric, Noise ratio=50%2024.08 | 93.9 | — | |
| CSBackbone=VGG-16, Subnetwork Type=Best Performing, Round=42019.12 | 93.45 | 2.4 | |
| DivideMixNoise type=Assymetric, Noise ratio=40%2024.08 | 93.4 | — | |
| SelCL+Noise type=Assymetric, Noise ratio=40%2024.08 | 93.4 | — | |
| CSBackbone=VGG-16, Subnetwork Type=Sparsest Matching, Round=52019.12 | 93.35 | 1.7 | |
| ELR+Noise type=Symmetric, Noise ratio=80%2024.08 | 93.3 | — | |
| DivideMixNoise type=Symmetric, Noise ratio=80%2024.08 | 93.2 | — | |
| MOITNoise type=Symmetric, Noise ratio=20%2024.08 | 93.1 | — | |
| ELR+Noise type=Assymetric, Noise ratio=40%2024.08 | 93 | — | |
| IMPBackbone=VGG-16, Subnetwork Type=Best Performing, Round=132019.12 | 92.97 | 5.5 | |
| IMP-CBackbone=VGG-16, Subnetwork Type=Best Performing, Round=122019.12 | 92.77 | 6.9 | |
| TCLNoise type=Assymetric, Noise ratio=40%2024.08 | 92.6 | — | |
| IMP-CBackbone=VGG-16, Subnetwork Type=Sparsest Matching, Round=182019.12 | 92.56 | 1.8 | |
| TCLNoise type=Symmetric, Noise ratio=80%2024.08 | 92.5 | — | |
| PENCILNoise type=Symmetric, Noise ratio=20%2024.08 | 92.4 | — | |
| IMPBackbone=VGG-16, Subnetwork Type=Sparsest Matching, Round=182019.12 | 92.36 | 1.8 | |
| Dense NetworkBackbone=VGG-16, Round=12019.12 | 92.35 | 100 | |
| LossModellingNoise type=Symmetric, Noise ratio=50%2024.08 | 92 | — | |
| MOITNoise type=Assymetric, Noise ratio=40%2024.08 | 92 | — | |
| CSBackbone=ResNet-20, Subnetwork Type=Best Performing, Round=52019.12 | 91.54 | 16.9 | |
| CSBackbone=ResNet-20, Subnetwork Type=Sparsest Matching, Round=52019.12 | 91.43 | 12.3 | |
| IMP-CBackbone=ResNet-20, Subnetwork Type=Best Performing, Round=42019.12 | 91.08 | 40.9 | |
| IMP-CBackbone=ResNet-20, Subnetwork Type=Sparsest Matching, Round=82019.12 | 91 | 16.7 | |
| IMPBackbone=ResNet-20, Subnetwork Type=Best Performing, Round=62019.12 | 90.67 | 26.2 | |
| IMPBackbone=ResNet-20, Subnetwork Type=Sparsest Matching, Round=72019.12 | 90.57 | 20.9 | |
| Dense NetworkBackbone=ResNet-20, Round=12019.12 | 90.55 | 100 | |
| MOITNoise type=Symmetric, Noise ratio=50%2024.08 | 90 | — | |
| Co-teaching+Noise type=Symmetric, Noise ratio=20%2024.08 | 89.5 | — | |
| TCLNoise type=Symmetric, Noise ratio=90%2024.08 | 89.4 | — | |
| SelCL+Noise type=Symmetric, Noise ratio=80%2024.08 | 89.2 | — | |
| PENCILNoise type=Symmetric, Noise ratio=50%2024.08 | 89.1 | — | |
| PENCILNoise type=Assymetric, Noise ratio=40%2024.08 | 88.5 | — | |
| LossModellingNoise type=Assymetric, Noise ratio=40%2024.08 | 87.4 | — | |
| F-correctionNoise type=Assymetric, Noise ratio=40%2024.08 | 87.2 | — | |
| CENoise type=Symmetric, Noise ratio=20%2024.08 | 86.8 | — | |
| F-correctionNoise type=Symmetric, Noise ratio=20%2024.08 | 86.8 | — | |
| LossModellingNoise type=Symmetric, Noise ratio=80%2024.08 | 86.8 | — | |
| Co-teaching+Noise type=Symmetric, Noise ratio=50%2024.08 | 85.7 | — | |
| CENoise type=Assymetric, Noise ratio=40%2024.08 | 85 | — | |
| SelCL+Noise type=Symmetric, Noise ratio=90%2024.08 | 81.9 | — | |
| F-correctionNoise type=Symmetric, Noise ratio=50%2024.08 | 79.8 | — | |
| CENoise type=Symmetric, Noise ratio=50%2024.08 | 79.4 | — | |
| MOITNoise type=Symmetric, Noise ratio=80%2024.08 | 79 | — | |
| ELR+Noise type=Symmetric, Noise ratio=90%2024.08 | 78.7 | — | |
| PENCILNoise type=Symmetric, Noise ratio=80%2024.08 | 77.5 | — | |
| DivideMixNoise type=Symmetric, Noise ratio=90%2024.08 | 76 | — | |
| MOITNoise type=Symmetric, Noise ratio=90%2024.08 | 69.6 | — | |
| LossModellingNoise type=Symmetric, Noise ratio=90%2024.08 | 69.1 | — | |
| Co-teaching+Noise type=Symmetric, Noise ratio=80%2024.08 | 67.4 | — | |
| F-correctionNoise type=Symmetric, Noise ratio=80%2024.08 | 63.3 | — | |
| CENoise type=Symmetric, Noise ratio=80%2024.08 | 62.9 | — | |
| PENCILNoise type=Symmetric, Noise ratio=90%2024.08 | 58.9 | — | |
| Co-teaching+Noise type=Symmetric, Noise ratio=90%2024.08 | 47.9 | — | |
| F-correctionNoise type=Symmetric, Noise ratio=90%2024.08 | 42.9 | — | |
| CENoise type=Symmetric, Noise ratio=90%2024.08 | 42.7 | — |