Reinforcement Learning on Seaquest Atari 2600 (test)
28,010Avg Total RewardHuman
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Humanagent=expert human player2013.12 | 28,010 | — | |
| DQNpolicy=epsilon-greedy, epsilon=0.052013.12 | 1,705 | — | |
| Contingencypolicy=epsilon-greedy, epsilon=0.052013.12 | 723 | — | |
| Sarsapolicy=epsilon-greedy, epsilon=0.052013.12 | 665 | — | |
| Randompolicy=random2013.12 | 110 | — | |
| DQN Bestpolicy=epsilon-greedy, epsilon=0.05, evaluation=best episode2013.12 | — | 1,740 | |
| HNeat Bestpolicy=deterministic2013.12 | — | 920 | |
| HNeat Pixelpolicy=deterministic, input=8 color channel representation2013.12 | — | 800 |