Task-incremental learning on Split Tiny-ImageNet (test)
82.04Average Accuracy (A_T)MULTITASK
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MULTITASKLearning Scenario=Standard Continual Learning, Replay Buffer (B)=true2023.06 | 82.04 | — | |
| KD w/ BUFFERLearning Scenario=Standard Continual Learning, Replay Buffer (B)=true2023.06 | 69.07 | 4.37 | |
| KD w/ QDILearning Scenario=Standard Continual Learning, Replay Buffer (B)=false2023.06 | 66.79 | 7.28 | |
| KDLearning Scenario=Standard Continual Learning, Replay Buffer (B)=false2023.06 | 65.48 | 7.33 | |
| COLD-ORACLEMemory Size (M)=2k2026.06 | 56.74 | 8 | |
| REFRESH-2024Memory Size (M)=2k2026.06 | 56.68 | 6 | |
| DER++-2020Memory Size (M)=2k2026.06 | 56.19 | 5 | |
| CBA-2023Memory Size (M)=2k2026.06 | 55.92 | 4 | |
| COLDMemory Size (M)=2k2026.06 | 55.09 | 10 | |
| KD w/ BUFFERLearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=true2023.06 | 54.84 | 10.82 | |
| KD w/ QDILearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=false2023.06 | 53.39 | 13.66 | |
| KDLearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=false2023.06 | 52.86 | 11.73 | |
| ER-ACE-2022Memory Size (M)=2k2026.06 | 51.37 | 7 | |
| A-GEM-2019Memory Size (M)=2k2026.06 | 51.11 | 23 | |
| DILearning Scenario=Standard Continual Learning, Replay Buffer (B)=false2023.06 | 44.29 | 22.42 | |
| SILearning Scenario=Standard Continual Learning, Replay Buffer (B)=false2023.06 | 36.32 | — | |
| MER-2019Memory Size (M)=2k2026.06 | 34.89 | 5 | |
| DER++Learning Scenario=Standard Continual Learning, Replay Buffer (B)=true2023.06 | 33.63 | 40.66 | |
| DILearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=false2023.06 | 31.69 | 32.42 | |
| DERLearning Scenario=Standard Continual Learning, Replay Buffer (B)=true2023.06 | 30.97 | 45.25 | |
| DER++Learning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=true2023.06 | 29.74 | 36.79 | |
| DER-2020Memory Size (M)=2k2026.06 | 28.22 | 20 | |
| ICARLLearning Scenario=Standard Continual Learning, Replay Buffer (B)=true2023.06 | 28.19 | — | |
| ERLearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=true2023.06 | 27.82 | 39.66 | |
| DERLearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=true2023.06 | 25.27 | 40.81 | |
| EWC-2017Memory Size (M)=2k2026.06 | 24.12 | 12 | |
| A-GEMLearning Scenario=Standard Continual Learning, Replay Buffer (B)=true2023.06 | 22.44 | 55.43 | |
| ERLearning Scenario=Standard Continual Learning, Replay Buffer (B)=true2023.06 | 21.03 | 57.88 | |
| GDumb-2020Memory Size (M)=2k2026.06 | 21 | — | |
| NCCL-2023Memory Size (M)=2k2026.06 | 19.64 | 44 | |
| FINETUNELearning Scenario=Standard Continual Learning, Replay Buffer (B)=false2023.06 | 19.31 | 59.5 | |
| Fine-tuneMemory Size (M)=2k2026.06 | 16.59 | 44 | |
| LWFLearning Scenario=Standard Continual Learning, Replay Buffer (B)=false2023.06 | 15.85 | — | |
| A-GEMLearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=true2023.06 | 15.79 | 54.1 | |
| FINETUNELearning Scenario=Heterogeneous Continual Learning, Replay Buffer (B)=false2023.06 | 14.83 | 56.31 |