Class Incremental Learning on CIFAR100 Ns=5 (test)
80.41Accuracy (LA)Joint (ResNet18)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Joint (ResNet18)Backbone=ResNet182023.03 | 80.41 | 0 | 0 | — | |
| Joint (Transf.)Backbone=Vision Transformer2023.03 | 76.12 | 0 | 0 | — | |
| DNEBackbone=Vision Transformer, Heads per task=22023.03 | 69.72 | 74.27 | 6.4 | — | |
| DNEBackbone=Vision Transformer, Heads per task=42023.03 | 69.43 | 74.2 | 6.69 | — | |
| DNEBackbone=Vision Transformer, Heads per task=12023.03 | 69.1 | 74.03 | 7.02 | — | |
| DERBackbone=ResNet182023.03 | 60.73 | 70.42 | 15.39 | — | |
| DytoxBackbone=Vision Transformer2023.03 | 58.59 | 68.31 | 17.53 | — | |
| PODNetBackbone=ResNet182023.03 | 48.18 | 60.69 | 27.94 | — | |
| FOSTERBackbone=ResNet182023.03 | 48.18 | 60.69 | 27.94 | — | |
| iCaRLBackbone=ResNet182023.03 | 42.89 | 55.23 | 33.23 | — |