Task-incremental learning on CIFAR-10 Split (test)
98.31Average AccuracyMULTITASK
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MULTITASKLearning Scenario=Standard Continual Learning, Replay Buffer (B)=true2023.06 | 98.31 | — | |
| KD w/ BUFFERLearning Scenario=Standard Continual Learning, Replay Buffer (B)=true2023.06 | 95.81 | 0.73 | |
| KD w/ QDILearning Scenario=Standard Continual Learning, Replay Buffer (B)=false2023.06 | 95.46 | 0.37 | |
| KDLearning Scenario=Standard Continual Learning, Replay Buffer (B)=false2023.06 | 95.01 | 0.87 | |
| KD w/ BUFFERLearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=true2023.06 | 93.95 | 2.49 | |
| KDLearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=false2023.06 | 93.13 | 3.97 | |
| DERLearning Scenario=Standard Continual Learning, Replay Buffer (B)=true2023.06 | 92.45 | 5.79 | |
| DER++Learning Scenario=Standard Continual Learning, Replay Buffer (B)=true2023.06 | 92.16 | 5.96 | |
| DER++ w/ AdaB2NBase Algorithm=DER++, Normalization=AdaB2N, Batch Size |B|=10, Memory Size |M|=20002023.10 | 91.99 | — | |
| ERLearning Scenario=Standard Continual Learning, Replay Buffer (B)=true2023.06 | 91.39 | 5.99 | |
| ER-ACE w/ AdaB2NBase Algorithm=ER-ACE, Normalization=AdaB2N, Batch Size |B|=10, Memory Size |M|=20002023.10 | 90.84 | — | |
| DER++ w/ BNBase Algorithm=DER++, Normalization=BN, Batch Size |B|=10, Memory Size |M|=20002023.10 | 90.42 | — | |
| KD w/ QDILearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=false2023.06 | 90.3 | 6.34 | |
| DER++ w/ AdaB2NBase Algorithm=DER++, Normalization=AdaB2N, Batch Size |B|=10, Memory Size |M|=5002023.10 | 90.15 | — | |
| ER-ACE w/ CNBase Algorithm=ER-ACE, Normalization=CN, Batch Size |B|=10, Memory Size |M|=20002023.10 | 90.01 | — | |
| DER++ w/ CNBase Algorithm=DER++, Normalization=CN, Batch Size |B|=10, Memory Size |M|=20002023.10 | 89.75 | — | |
| ER-ACE w/ BNBase Algorithm=ER-ACE, Normalization=BN, Batch Size |B|=10, Memory Size |M|=20002023.10 | 89.17 | — | |
| DILearning Scenario=Standard Continual Learning, Replay Buffer (B)=false2023.06 | 89.11 | 7.12 | |
| ICARLLearning Scenario=Standard Continual Learning, Replay Buffer (B)=true2023.06 | 88.99 | 2.63 | |
| ER-ACE w/ AdaB2NBase Algorithm=ER-ACE, Normalization=AdaB2N, Batch Size |B|=10, Memory Size |M|=5002023.10 | 88.74 | — | |
| ER-ACE w/ CNBase Algorithm=ER-ACE, Normalization=CN, Batch Size |B|=10, Memory Size |M|=5002023.10 | 88.32 | — | |
| DER++ w/ BNBase Algorithm=DER++, Normalization=BN, Batch Size |B|=10, Memory Size |M|=5002023.10 | 87.61 | — | |
| A-GEMLearning Scenario=Standard Continual Learning, Replay Buffer (B)=true2023.06 | 87.09 | 12.54 | |
| DER++ w/ INBase Algorithm=DER++, Normalization=IN, Batch Size |B|=10, Memory Size |M|=20002023.10 | 87.04 | — | |
| DER++ w/ CNBase Algorithm=DER++, Normalization=CN, Batch Size |B|=10, Memory Size |M|=5002023.10 | 86.92 | — | |
| ER-ACE w/ BNBase Algorithm=ER-ACE, Normalization=BN, Batch Size |B|=10, Memory Size |M|=5002023.10 | 86.34 | — | |
| ER-ACE w/ INBase Algorithm=ER-ACE, Normalization=IN, Batch Size |B|=10, Memory Size |M|=20002023.10 | 86.03 | — | |
| COLD-ORACLEMemory Size (M)=6002026.06 | 85.68 | 6 | |
| CBA-2023Memory Size (M)=6002026.06 | 85.01 | 2 | |
| REFRESH-2024Memory Size (M)=6002026.06 | 84.93 | 3 | |
| COLDMemory Size (M)=6002026.06 | 84.9 | 7 | |
| DER++-2020Memory Size (M)=6002026.06 | 84.84 | 3 | |
| DER++ w/ INBase Algorithm=DER++, Normalization=IN, Batch Size |B|=10, Memory Size |M|=5002023.10 | 84.81 | — | |
| ER-ACE w/ INBase Algorithm=ER-ACE, Normalization=IN, Batch Size |B|=10, Memory Size |M|=5002023.10 | 84.57 | — | |
| A-GEM-2019Memory Size (M)=6002026.06 | 83.75 | 42 | |
| DER-2020Memory Size (M)=6002026.06 | 83.66 | 5 | |
| DER++ w/ GNBase Algorithm=DER++, Normalization=GN, Batch Size |B|=10, Memory Size |M|=20002023.10 | 83.34 | — | |
| ER-ACE w/ GNBase Algorithm=ER-ACE, Normalization=GN, Batch Size |B|=10, Memory Size |M|=20002023.10 | 82.85 | — | |
| DER++ w/ LNBase Algorithm=DER++, Normalization=LN, Batch Size |B|=10, Memory Size |M|=20002023.10 | 82.8 | — | |
| ER-ACE-2022Memory Size (M)=6002026.06 | 82.7 | 3 | |
| DER++ w/ GNBase Algorithm=DER++, Normalization=GN, Batch Size |B|=10, Memory Size |M|=5002023.10 | 82.05 | — | |
| ERLearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=true2023.06 | 81.78 | 17.57 | |
| ER-ACE w/ LNBase Algorithm=ER-ACE, Normalization=LN, Batch Size |B|=10, Memory Size |M|=20002023.10 | 81.59 | — | |
| ER-ACE w/ GNBase Algorithm=ER-ACE, Normalization=GN, Batch Size |B|=10, Memory Size |M|=5002023.10 | 81.46 | — | |
| DER++ w/ LNBase Algorithm=DER++, Normalization=LN, Batch Size |B|=10, Memory Size |M|=5002023.10 | 81.35 | — | |
| DER++Learning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=true2023.06 | 81.33 | 18.79 | |
| DILearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=false2023.06 | 81.16 | 15.84 | |
| DERLearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=true2023.06 | 79.92 | 19.99 | |
| ER-ACE w/ LNBase Algorithm=ER-ACE, Normalization=LN, Batch Size |B|=10, Memory Size |M|=5002023.10 | 78.01 | — | |
| MER-2019Memory Size (M)=6002026.06 | 77.32 | 4 | |
| A-GEMLearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=true2023.06 | 76.7 | 24.31 | |
| Fine-tuneMemory Size (M)=6002026.06 | 76.7 | 22 | |
| NCCL-2023Memory Size (M)=6002026.06 | 72.99 | 25 | |
| EWC-2017Memory Size (M)=6002026.06 | 71.05 | 35 | |
| SILearning Scenario=Standard Continual Learning, Replay Buffer (B)=false2023.06 | 68.05 | 38.76 | |
| GDumb-2020Memory Size (M)=6002026.06 | 64.68 | — | |
| LWFLearning Scenario=Standard Continual Learning, Replay Buffer (B)=false2023.06 | 63.29 | 32.56 | |
| FINETUNELearning Scenario=Standard Continual Learning, Replay Buffer (B)=false2023.06 | 62.93 | 41.45 | |
| FINETUNELearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=false2023.06 | 61.99 | 42.15 |