Image Classification on CIFAR-10 original (test)
98.13AccuracyVanilla
Evaluation results
| Method | Links | |
|---|---|---|
| VanillaBackbone=ViT-Base, Pre-training=ImageNet-1K2023.10 | 98.13 | |
| VanillaBackbone=ResNet50, Pre-training=ImageNet-1K2023.10 | 96.97 | |
| C8-Aug.Backbone=ViT-Base, Pre-training=ImageNet-1K2023.10 | 96.36 | |
| Rotation AugmentationBackbone=ViT-Base, Pre-training=ImageNet-1K2023.10 | 96.26 | |
| Prior-Regularized LCBackbone=ResNet50, Pre-training=ImageNet-1K2023.10 | 96.19 | |
| Prior-Regularized LCBackbone=ViT-Base, Pre-training=ImageNet-1K2023.10 | 96.14 | |
| C8-Aug.Backbone=ResNet50, Pre-training=ImageNet-1K2023.10 | 95.76 | |
| MutexMatchNumber of labels=80 labels, k=0.6C2022.03 | 95 | |
| Learned Canonicalization (LC)Backbone=ViT-Base, Pre-training=ImageNet-1K2023.10 | 95 | |
| SLANumber of labels=80 labels2022.03 | 94.98 | |
| Rotation AugmentationBackbone=ResNet50, Pre-training=ImageNet-1K2023.10 | 94.91 | |
| SLANumber of labels=40 labels2022.03 | 94.83 | |
| MutexMatchNumber of labels=40 labels, k=0.6C2022.03 | 94.21 | |
| CoMatchNumber of labels=80 labels, Distribution Alignment=true2022.03 | 94.08 | |
| Learned Canonicalization (LC)Backbone=ResNet50, Pre-training=ImageNet-1K2023.10 | 93.29 | |
| Scratch-BModel=VGG-16, Pruning=l1-norm based filter pruning2021.05 | 93.24 | |
| MutexMatchNumber of labels=80 labels, Distribution Alignment=true2022.03 | 93.23 | |
| MutexMatchNumber of labels=40 labels, Distribution Alignment=true2022.03 | 93.22 | |
| CoMatchNumber of labels=40 labels, Distribution Alignment=true2022.03 | 93.21 | |
| Scratch-BModel=ResNet-110, Version=B, Pruning=l1-norm based filter pruning2021.05 | 93.2 | |
| Cyclic Learning Rate Restarting (CLR)Model=VGG-16, Pruning=l1-norm based filter pruning2021.05 | 93.12 | |
| Fine-tuningModel=VGG-16, Pruning=l1-norm based filter pruning2021.05 | 93.09 | |
| Cyclic Learning Rate Restarting (CLR)Model=ResNet-110, Version=A, Pruning=l1-norm based filter pruning2021.05 | 93.08 | |
| Scratch-BModel=ResNet-110, Version=A, Pruning=l1-norm based filter pruning2021.05 | 93.04 | |
| Cyclic Learning Rate Restarting (CLR)Model=ResNet-110, Version=B, Pruning=l1-norm based filter pruning2021.05 | 93.03 | |
| Unpruned BaselineModel=VGG-162021.05 | 93.01 | |
| Scratch-EModel=VGG-16, Pruning=l1-norm based filter pruning2021.05 | 93.01 | |
| Scratch-EModel=ResNet-110, Version=A, Pruning=l1-norm based filter pruning2021.05 | 92.97 | |
| Unpruned BaselineModel=ResNet-1102021.05 | 92.91 | |
| Fine-tuningModel=ResNet-110, Version=A, Pruning=l1-norm based filter pruning2021.05 | 92.81 | |
| Scratch-BModel=ResNet-56, Version=A, Pruning=l1-norm based filter pruning2021.05 | 92.77 | |
| Cyclic Learning Rate Restarting (CLR)Model=ResNet-56, Version=A, Pruning=l1-norm based filter pruning2021.05 | 92.71 | |
| Scratch-EModel=ResNet-110, Version=B, Pruning=l1-norm based filter pruning2021.05 | 92.63 | |
| Scratch-EModel=ResNet-56, Version=A, Pruning=l1-norm based filter pruning2021.05 | 92.6 | |
| Cyclic Learning Rate Restarting (CLR)Model=ResNet-56, Version=B, Pruning=l1-norm based filter pruning2021.05 | 92.41 | |
| Scratch-EModel=ResNet-56, Version=B, Pruning=l1-norm based filter pruning2021.05 | 92.39 | |
| Fine-tuningModel=ResNet-110, Version=B, Pruning=l1-norm based filter pruning2021.05 | 92.34 | |
| Unpruned BaselineModel=ResNet-562021.05 | 92.32 | |
| Scratch-BModel=ResNet-56, Version=B, Pruning=l1-norm based filter pruning2021.05 | 92.31 | |
| MutexMatchNumber of labels=20 labels, k=0.6C2022.03 | 92.23 | |
| Fine-tuningModel=ResNet-56, Version=A, Pruning=l1-norm based filter pruning2021.05 | 92.18 | |
| Fine-tuningModel=ResNet-56, Version=B, Pruning=l1-norm based filter pruning2021.05 | 92.06 | |
| FixMatchNumber of labels=80 labels2022.03 | 91.99 | |
| MutexMatchNumber of labels=20 labels, Distribution Alignment=true2022.03 | 91.77 | |
| FixMatchNumber of labels=40 labels2022.03 | 89.18 | |
| CoMatchNumber of labels=20 labels, Distribution Alignment=true2022.03 | 88.43 | |
| FixMatchNumber of labels=20 labels2022.03 | 84.97 | |
| CFMIPC=50, Backbone=ResNet-182025.03 | 82.3 | |
| SLANumber of labels=20 labels2022.03 | 81.91 | |
| ReMixMatchNumber of labels=40 labels, Distribution Alignment=true2022.03 | 80.9 | |
| MutexMatchNumber of labels=10 labels, Distribution Alignment=true2022.03 | 76.06 | |
| FrePoIPC=50, Backbone=ConvNet-1282025.03 | 71.7 | |
| MTTIPC=50, Backbone=ConvNet-1282025.03 | 71.6 | |
| UDANumber of labels=40 labels2022.03 | 70.95 | |
| CoMatchNumber of labels=10 labels, Distribution Alignment=true2022.03 | 69.87 | |
| MutexMatchNumber of labels=10 labels, Distribution Alignment=false2022.03 | 66.45 | |
| SLANumber of labels=10 labels2022.03 | 65.87 | |
| FrePoIPC=10, Backbone=ConvNet-1282025.03 | 65.5 | |
| MTTIPC=10, Backbone=ConvNet-1282025.03 | 65.3 | |
| CFMIPC=50, Backbone=ConvNet-1282025.03 | 64 | |
| DWAIPC=50, Backbone=ConvNet-1282025.03 | 63.3 | |
| DMIPC=50, Backbone=ConvNet-1282025.03 | 63 | |
| G-VBDMIPC=50, Backbone=ResNet-182025.03 | 59.2 | |
| MutexMatchNumber of labels=10 labels, k=0.6C2022.03 | 57.52 | |
| CFMIPC=10, Backbone=ResNet-182025.03 | 57 | |
| G-VBSMIPC=50, Backbone=ConvNet-1282025.03 | 54.3 | |
| DCIPC=50, Backbone=ConvNet-1282025.03 | 53.9 | |
| G-VBDMIPC=10, Backbone=ResNet-182025.03 | 53.5 | |
| DWAIPC=50, Backbone=ResNet-182025.03 | 53.1 | |
| MixMatchNumber of labels=40 labels2022.03 | 52.46 | |
| CFMIPC=10, Backbone=ConvNet-1282025.03 | 52.1 | |
| DMIPC=10, Backbone=ConvNet-1282025.03 | 48.9 | |
| SRe2LIPC=50, Backbone=ResNet-182025.03 | 47.5 | |
| G-VBSMIPC=10, Backbone=ConvNet-1282025.03 | 46.5 | |
| FixMatchNumber of labels=10 labels2022.03 | 45.91 | |
| FedCILLocal training iteration (T)=400, Global communication round (R)=200, Mini-batch size (B)=100, Backbone=3-layer CNN, Optimizer=Adam, Learning rate=1e-42023.02 | 45.27 | |
| DWAIPC=10, Backbone=ConvNet-1282025.03 | 45 | |
| DCIPC=10, Backbone=ConvNet-1282025.03 | 44.9 | |
| FedProx+ACGAN ReplayLocal training iteration (T)=400, Global communication round (R)=200, Mini-batch size (B)=100, Backbone=3-layer CNN, Optimizer=Adam, Learning rate=1e-42023.02 | 38.34 | |
| FedAvg+ACGAN ReplayLocal training iteration (T)=400, Global communication round (R)=200, Mini-batch size (B)=100, Backbone=3-layer CNN, Optimizer=Adam, Learning rate=1e-42023.02 | 38.31 | |
| FedAvg+DGRLocal training iteration (T)=400, Global communication round (R)=200, Mini-batch size (B)=100, Backbone=3-layer CNN, Optimizer=Adam, Learning rate=1e-42023.02 | 37.93 | |
| FedProx+DGRLocal training iteration (T)=400, Global communication round (R)=200, Mini-batch size (B)=100, Backbone=3-layer CNN, Optimizer=Adam, Learning rate=1e-42023.02 | 37.87 | |
| DWAIPC=10, Backbone=ResNet-182025.03 | 32.6 | |
| FedProxLocal training iteration (T)=400, Global communication round (R)=200, Mini-batch size (B)=100, Backbone=3-layer CNN, Optimizer=Adam, Learning rate=1e-42023.02 | 27.43 | |
| FedAvgLocal training iteration (T)=400, Global communication round (R)=200, Mini-batch size (B)=100, Backbone=3-layer CNN, Optimizer=Adam, Learning rate=1e-42023.02 | 27.21 | |
| SRe2LIPC=10, Backbone=ResNet-182025.03 | 27.2 | |
| FedLwF-2TLocal training iteration (T)=400, Global communication round (R)=200, Mini-batch size (B)=100, Backbone=3-layer CNN, Optimizer=Adam, Learning rate=1e-42023.02 | 27.02 |