Class-incremental Continual Learning on Tiny-ImageNet Split (test)
59.99Average Accuracy (AT)MULTITASK
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MULTITASKProtocol=Standard Continual Learning, Replay-buffer (B)=false2023.06 | 59.99 | — | |
| KD w/ BUFFERProtocol=Standard Continual Learning, Replay-buffer (B)=true2023.06 | 29.21 | 16.48 | |
| KD w/ QDIProtocol=Standard Continual Learning, Replay-buffer (B)=false2023.06 | 24.21 | 36.58 | |
| KDProtocol=Standard Continual Learning, Replay-buffer (B)=false2023.06 | 20.86 | 42.38 | |
| KD w/ BUFFERProtocol=Heterogeneous Continual Learning, Replay-buffer (B)=true2023.06 | 18.77 | 8.76 | |
| KD w/ QDIProtocol=Heterogeneous Continual Learning, Replay-buffer (B)=false2023.06 | 15.38 | 37.27 | |
| KDProtocol=Heterogeneous Continual Learning, Replay-buffer (B)=false2023.06 | 14.18 | 46.91 | |
| ERProtocol=Standard Continual Learning, Replay-buffer (B)=true2023.06 | 10.7 | 83.99 | |
| DERProtocol=Standard Continual Learning, Replay-buffer (B)=true2023.06 | 10.01 | 65.66 | |
| DER++Protocol=Heterogeneous Continual Learning, Replay-buffer (B)=true2023.06 | 8.88 | 55.54 | |
| LwFProtocol=Standard Continual Learning, Replay-buffer (B)=false2023.06 | 8.46 | — | |
| DER++Protocol=Standard Continual Learning, Replay-buffer (B)=true2023.06 | 8.23 | 68.51 | |
| DERProtocol=Heterogeneous Continual Learning, Replay-buffer (B)=true2023.06 | 8.11 | 56.73 | |
| FINETUNEProtocol=Heterogeneous Continual Learning, Replay-buffer (B)=false2023.06 | 7.9 | 57.23 | |
| ERProtocol=Heterogeneous Continual Learning, Replay-buffer (B)=true2023.06 | 7.59 | 61.77 | |
| iCaRLProtocol=Standard Continual Learning, Replay-buffer (B)=true2023.06 | 7.53 | — | |
| A-GEMProtocol=Standard Continual Learning, Replay-buffer (B)=true2023.06 | 7.49 | 72.04 | |
| DIProtocol=Standard Continual Learning, Replay-buffer (B)=false2023.06 | 7.28 | 51.7 | |
| A-GEMProtocol=Heterogeneous Continual Learning, Replay-buffer (B)=true2023.06 | 6.93 | 63.93 | |
| FINETUNEProtocol=Standard Continual Learning, Replay-buffer (B)=false2023.06 | 6.78 | 40.67 | |
| SIProtocol=Standard Continual Learning, Replay-buffer (B)=false2023.06 | 6.58 | — | |
| DIProtocol=Heterogeneous Continual Learning, Replay-buffer (B)=false2023.06 | 6.39 | 51.14 |