Continual Learning on Spatial task benchmark
95Average AccuracyJoint oracle
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Joint oracleSeeds=102026.06 | 95 | — | |
| ReplayStores raw data?=yes, Samples per task=200, Seeds=102026.06 | 94 | 5.6 | |
| Stefan-CLStores raw data?=no, Seeds=102026.06 | 92.3 | 2.1 | |
| EWCStores raw data?=no, lambda=300, Seeds=102026.06 | 71.6 | 28.7 | |
| SIStores raw data?=no, lambda=50, Seeds=102026.06 | 70.1 | 24.1 | |
| Naive sequentialStores raw data?=no, Seeds=102026.06 | 51.4 | 60.3 |