Generative Modeling on Sequential ImageNet (Cats), CUB 200, LSUN (Churches & Towers)
29FID (Cats, Task 1)Finetune
Evaluation results
| Method | Links | |||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Finetune2021.03 | 29 | 156.9 | 189.6 | 182.8 | — | 21.2 | 174.5 | 161.5 | — | — | 11.4 | 48 | — | — | — | 12.7 | 101.2 | |
| EWC2021.03 | 29 | 147.3 | 190.7 | 186.2 | — | 65.9 | 165.4 | 155.9 | — | — | 38.2 | 48.2 | — | — | — | 33.8 | 106.1 | |
| MeRGAN-RA2021.03 | 29 | 56.4 | 58.2 | 61.3 | — | 50.9 | 53.7 | 65.9 | — | — | 23.2 | 28.3 | — | — | — | 15.7 | 42.8 | |
| EFTrank configuration=a=32, b=16, parameter expansion=+4.8%2021.03 | 29 | 29 | 29 | 29 | — | 44.1 | 44.1 | 44.1 | — | — | 32.3 | 32.3 | — | — | — | 27.2 | 33.1 |