Reinforcement Learning on Hopper v5 (strong-drift)
20.24Final ReturnGB-DQN
Evaluation results
| Method | Links | |
|---|---|---|
| GB-DQNshrinkage coefficient η_boost=0.1, buffer capacity=200,0002025.12 | 20.24 | |
| DQNbuffer capacity=200,0002025.12 | 19.9 | |
| Ensemble-DQNbuffer capacity=200,0002025.12 | 18.14 | |
| Sliding-DQNbuffer capacity=50,0002025.12 | 17.1 | |
| Reset-DQNbuffer capacity=200,0002025.12 | 16.32 |