Image Classification on CIFAR100 LT
62.69Balanced AccuracyCross-Entropy
Evaluation results
| Method | Links | |
|---|---|---|
| Cross-Entropy2022.01 | 62.69 | |
| FixMatch + LA + DASOImbalance ratio (gamma)=10, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 60.6 | |
| FixMatch + ABC + DASOImbalance ratio (gamma)=10, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 60 | |
| FixMatch + LA + DARPImbalance ratio (gamma)=10, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 59.9 | |
| LDAM loss2022.01 | 59.47 | |
| FixMatch + DASOImbalance ratio (gamma)=10, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 59.2 | |
| FixMatch + ABCImbalance ratio (gamma)=10, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 59.1 | |
| LA lossτ=12022.01 | 58.96 | |
| AutoBalance (Algo. 1)Design Space=α ← ι2022.01 | 58.71 | |
| Supervised + LAImbalance ratio (gamma)=10, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 58.7 | |
| AutoBalance (Algo. 1)Design Space=α ← γ of LA loss2022.01 | 58.68 | |
| FixMatch + LAImbalance ratio (gamma)=10, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 58.6 | |
| AutoBalance (Algo. 1)Design Space=α ← Δ2022.01 | 58.4 | |
| FixMatch + DARPImbalance ratio (gamma)=10, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 58.1 | |
| FixMatch + CReST+Imbalance ratio (gamma)=10, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 57.4 | |
| CDT loss2022.01 | 57.26 | |
| FixMatch + LA + CReST+Imbalance ratio (gamma)=10, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 57.1 | |
| AutoBalance (Algo. 1)Design Space=α ← Δ & ι2022.01 | 56.84 | |
| AutoBalance (Algo. 1)Design Space=α ← Δ & ι, Initialization=LA init2022.01 | 56.7 | |
| FixMatchImbalance ratio (gamma)=10, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 56.5 | |
| FixMatch + ABC + DASOImbalance ratio (gamma)=20, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 55.3 | |
| FixMatch + LA + DASOImbalance ratio (gamma)=20, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 55.1 | |
| FixMatch + LA + DARPImbalance ratio (gamma)=20, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 53.8 | |
| FixMatch + ABCImbalance ratio (gamma)=20, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 53.7 | |
| FixMatch + LAImbalance ratio (gamma)=20, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 53.4 | |
| FixMatch + DASOImbalance ratio (gamma)=20, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 52.9 | |
| FixMatch + LA + CReST+Imbalance ratio (gamma)=20, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 52.3 | |
| FixMatch + DARPImbalance ratio (gamma)=20, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 52.2 | |
| FixMatch + CReST+Imbalance ratio (gamma)=20, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 52.1 | |
| FixMatchImbalance ratio (gamma)=20, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 50.7 | |
| FixMatch + LA + DASOImbalance ratio (gamma)=10, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 50.7 | |
| FixMatch + LA + DARPImbalance ratio (gamma)=10, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 50.5 | |
| FixMatch + ABC + DASOImbalance ratio (gamma)=10, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 50.2 | |
| FixMatch + DASOImbalance ratio (gamma)=10, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 49.8 | |
| FixMatch + DARPImbalance ratio (gamma)=10, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 49.4 | |
| SupervisedImbalance ratio (gamma)=10, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 48.7 | |
| FixMatch + ABCImbalance ratio (gamma)=10, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 47.5 | |
| FixMatch + LAImbalance ratio (gamma)=10, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 47.3 | |
| FixMatchImbalance ratio (gamma)=10, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 45.2 | |
| FedBBSkewness (δ)=0.5, Number of clients (K)=16, Local training epochs=20, Model aggregation epochs=302026.06 | 45.13 | |
| FixMatch + CReST+Imbalance ratio (gamma)=10, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 44.5 | |
| FixMatch + ABC + DASOImbalance ratio (gamma)=20, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 44.5 | |
| FixMatch + LA + DARPImbalance ratio (gamma)=20, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 44.4 | |
| Supervised + LAImbalance ratio (gamma)=20, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 44.1 | |
| FixMatch + LA + DASOImbalance ratio (gamma)=20, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 44.1 | |
| FixMatch + LA + CReST+Imbalance ratio (gamma)=10, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 44 | |
| FixMatch + DASOImbalance ratio (gamma)=20, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 43.6 | |
| FixMatch + DARPImbalance ratio (gamma)=20, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 43.4 | |
| FixMatch + ABCImbalance ratio (gamma)=20, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 41.6 | |
| FixMatch + LAImbalance ratio (gamma)=20, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 41.4 | |
| FedAlignSkewness (δ)=0.5, Number of clients (K)=16, Local training epochs=20, Model aggregation epochs=302026.06 | 41.37 | |
| SupervisedImbalance ratio (gamma)=20, Number of labeled samples (N_l)=150, Number of unlabeled samples (M_u)=3002021.06 | 41.2 | |
| FedLCSkewness (δ)=0.5, Number of clients (K)=16, Local training epochs=20, Model aggregation epochs=302026.06 | 41.08 | |
| MOONSkewness (δ)=0.5, Number of clients (K)=16, Local training epochs=20, Model aggregation epochs=302026.06 | 41.02 | |
| FixMatch + LA + CReST+Imbalance ratio (gamma)=20, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 40.6 | |
| FedProxSkewness (δ)=0.5, Number of clients (K)=16, Local training epochs=20, Model aggregation epochs=302026.06 | 40.18 | |
| FixMatch + CReST+Imbalance ratio (gamma)=20, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 40.1 | |
| FedBBSkewness (δ)=0.1, Number of clients (K)=16, Local training epochs=20, Model aggregation epochs=302026.06 | 40.05 | |
| FixMatchImbalance ratio (gamma)=20, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 40 | |
| FedAvgSkewness (δ)=0.5, Number of clients (K)=16, Local training epochs=20, Model aggregation epochs=302026.06 | 39.75 | |
| FedAlignSkewness (δ)=0.1, Number of clients (K)=16, Local training epochs=20, Model aggregation epochs=302026.06 | 39.27 | |
| FedNovaSkewness (δ)=0.5, Number of clients (K)=16, Local training epochs=20, Model aggregation epochs=302026.06 | 38.98 | |
| MOONSkewness (δ)=0.1, Number of clients (K)=16, Local training epochs=20, Model aggregation epochs=302026.06 | 36.78 | |
| FedNovaSkewness (δ)=0.1, Number of clients (K)=16, Local training epochs=20, Model aggregation epochs=302026.06 | 36.72 | |
| FedAvgSkewness (δ)=0.1, Number of clients (K)=16, Local training epochs=20, Model aggregation epochs=302026.06 | 35.52 | |
| FedLCSkewness (δ)=0.1, Number of clients (K)=16, Local training epochs=20, Model aggregation epochs=302026.06 | 35.51 | |
| FedProxSkewness (δ)=0.1, Number of clients (K)=16, Local training epochs=20, Model aggregation epochs=302026.06 | 35.01 | |
| Supervised + LAImbalance ratio (gamma)=10, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 30.2 | |
| SupervisedImbalance ratio (gamma)=10, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 29.6 | |
| Supervised + LAImbalance ratio (gamma)=20, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 26.5 | |
| SupervisedImbalance ratio (gamma)=20, Number of labeled samples (N_l)=50, Number of unlabeled samples (M_u)=4002021.06 | 25.1 |