Lifelong Learning on 5-dataset
70AccuracyMTL
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MTLsetting=upper bound2021.05 | 70 | — | — | |
| TAG-RMSProp2021.05 | 62.59 | 0.29 | 86.08 | |
| ERepisodic memory size=1 example per class2021.05 | 61.58 | 0.28 | 84.3 | |
| A-GEMepisodic memory size=1 example per class2021.05 | 55.9 | 0.34 | 82.61 | |
| EWC2021.05 | 48.58 | 0.4 | 79.56 | |
| Stable SGD2021.05 | 46.51 | 0.46 | 83.3 | |
| Naive SGD2021.05 | 46.48 | 0.48 | 84.55 | |
| Naive RMSProp2021.05 | 45.49 | 0.5 | 85.58 |