Offline Reinforcement Learning on RL Unplugged Atari 46 games
265.3Median Human Normalized ScoreMuZero Unplugged
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MuZero UnpluggedArchitecture=MuZero network architecture, Loss=Reanalyse2021.04 | 265.3 | 595.5 | |
| CRRLoss=CRR2021.04 | 155.6 | 271.2 | |
| REMLoss=REM2021.04 | 107.9 | 113.5 | |
| BCQLoss=BCQ2021.04 | 107.5 | 120 | |
| IQNLoss=IQN2021.04 | 100.8 | 96.1 | |
| DQNLoss=DQN2021.04 | 86.2 | 89.5 | |
| MuZero BCArchitecture=MuZero network architecture, Loss=BC2021.04 | 54 | 46.9 | |
| BCLoss=BC2021.04 | 53.3 | 48.5 |