Continual Learning on PMNIST (test)
98.6AccuracyHAT - LARGE
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| HAT - LARGEPARAMETERS=5.8 M2018.01 | 98.6 | — | — | |
| HAT - MEDIUMPARAMETERS=0.7 M2018.01 | 97.4 | — | — | |
| SIPARAMETERS=5.8 M2018.01 | 97.1 | — | — | |
| EWCPARAMETERS=5.8 M2018.01 | 96.9 | — | — | |
| CABPARAMETERS=0.7 M2018.01 | 95.2 | — | — | |
| GENERATIVE REPLAYPARAMETERS=UNKNOWN*2018.01 | 94.9 | — | — | |
| HAT - SMALLPARAMETERS=0.1 M2018.01 | 91.6 | — | — | |
| VCLPARAMETERS=0.1 M*2018.01 | 90 | — | — | |
| MBPA + EWC – 1000 EX.PARAMETERS=UNKNOWN*2018.01 | 89.7 | — | — | |
| EWCPARAMETERS=0.1 M2018.01 | 88.2 | — | — | |
| SIPARAMETERS=0.1 M2018.01 | 86 | — | — | |
| Offlinetraining_mode=offline2022.09 | 84.95 | — | — | |
| GEMPARAMETERS=0.1 M*2018.01 | 82.8 | — | — | |
| CTN (+BOME)framework=CTN, optimizer=BOME2022.09 | 80.7 | 4.09 | 84.79 | |
| CTN (+ITD)framework=CTN, optimizer=ITD2022.09 | 78.4 | 5.62 | 84.02 | |
| CTN (+BVFSM)framework=CTN, optimizer=BVFSM2022.09 | 77.78 | 7.25 | 85.03 | |
| MERalgorithm=MER2022.09 | 76.59 | 5.73 | 82.32 |