Reinforcement Learning on Atari 10 hard-exploration games (train)
124,524.02Alien ScoreBYOL-Explore
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| BYOL-Exploreepisodes=10, seeds=32022.06 | 124,524.02 | 32.12 | 17,014.51 | 153,124.29 | 13,518.18 | 25,175.53 | 33,130.85 | 358,269.62 | 11,060.24 | 1,980.24 | |
| BYOL-Explore (Horizon=1)horizon=12022.06 | 105,384.32 | 32.73 | 13,621.41 | 67,954.99 | 6,701.77 | 17,363.46 | 95,359.23 | 368,757.5 | 9,585.28 | 2,078.24 | |
| BYOL-Explore (No clipping)clipping=false2022.06 | 99,570.34 | 30.13 | 14,949.85 | 55,897.06 | 5,093.32 | 19,427.77 | 31,048.56 | 370,719.6 | 10,380.94 | 2,121.38 | |
| BYOL-Explore (big)episodes=10, seeds=3, model_capacity=big2022.06 | 87,754.33 | 32.62 | 12,234.71 | 54,139.73 | 5,615.36 | 27,555.12 | 94,589.47 | 377,349.65 | 19,277.35 | 2,250.12 | |
| BYOL-Explore (No sharing)parameter_sharing=false2022.06 | 75,876.95 | 30.63 | 12,026.79 | 67,555.39 | 9,090.71 | 0 | 36,183.65 | 338,709.48 | 7,105.23 | 1,977.89 | |
| RNDepisodes=10, seeds=32022.06 | 74,717.02 | 33.74 | 5,885.23 | 36,452.78 | 4,554.37 | 0 | 33,150.09 | 34,341.06 | 5,262.55 | 1,879.68 | |
| RLepisodes=10, seeds=32022.06 | 18,252.69 | 33.97 | 13,541.1 | 37,235.84 | 160.11 | 0 | 5,704.99 | 91,920.56 | 8,147.67 | 670.61 | |
| Average Human2022.06 | 7,127.7 | 29.6 | 3,351.4 | 30,826.4 | 4,753.3 | 6,463.7 | 69,571.3 | 13,455 | 12,326.7 | 1,187.5 | |
| ICMepisodes=10, seeds=32022.06 | 5,664.83 | 32.61 | 4,657.2 | 18,753.97 | 300 | 0 | 1,419.49 | 44,294.63 | 3,924.22 | 1,860.19 | |
| BYOL-Explore (Fixed targets)target_network=fixed2022.06 | 411.63 | 6.09 | 401.79 | 1,963.1 | 8.52 | -3.55 | 1,706.72 | 565.59 | 3,772.82 | 31.12 |