Reinforcement Learning on Atari 2600 (Full Game Set Scores)
1,510.8Alien ScorePOP3D
Evaluation results
| Method | Links | |||||||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| POP3Devaluation_window=last 100 episodes, frame_count=40M, trials=32018.07 | 1,510.8 | 729.15 | 5,400.13 | 4,310.67 | 2,488.1 | 2,193,605.67 | 1,212.23 | 15,466.67 | 4,549 | 38.99 | 97.23 | 458.41 | 3,315.44 | 6,308.33 | 120,247.33 | 61,147.33 | -7.89 | 459.85 | 28.99 | 21.21 | 316.87 | 6,207 | 557.17 | -4.12 | 527.17 | 3,891.67 | 7,715.68 | 33,728 | 0 | 1,683.87 | 6,065.63 | 0 | 20.5 | 79.67 | 15,396.67 | 8,052.23 | 44,679.67 | 4.6 | 1,807.47 | 1,216.15 | 48,984 | -8.32 | 3,770.33 | 241.21 | 242,701.51 | 36.33 | 37,780.7 | 4,704 | 9,472 | |
| PPOevaluation_window=last 100 episodes, frame_count=40M, trials=32018.07 | 1,431.17 | 790.75 | 4,438.82 | 3,483.17 | 1,605.33 | 2,140,536.33 | 1,206.67 | 14,766.67 | 2,624.19 | 47.27 | 93.7 | 281.93 | 3,565.18 | 4,872.67 | 105,940 | 26,740.57 | -11.22 | 698.46 | 17.72 | 21.11 | 280.3 | 1,791 | 753.5 | -4.83 | 488.17 | 6,845 | 8,329.08 | 29,958.67 | 10.67 | 1,981.5 | 5,397.47 | -2.32 | 20.8 | 36.5 | 14,556.83 | 7,360.4 | 36,289.33 | 14.15 | 1,470.6 | 944.63 | 33,862 | -13.74 | 5,321.33 | 177.58 | 153,160.66 | 0 | 31,577.24 | 4,886.67 | 5,728.67 | |
| BASELINEevaluation_window=last 100 episodes, frame_count=40M, trials=3, mode=fixed KLD2018.07 | 1,311.23 | 655.1 | 1,846.75 | 3,657.67 | 1,615.37 | 1,515,993.33 | 1,124.43 | 14,690 | 6,898.09 | 30.48 | 65.33 | 67.7 | 3,393.93 | 2,676 | 98,219.67 | 57,476.65 | -8.61 | 518.41 | -64.27 | 18.37 | 280.3 | 940.87 | 449 | -3.61 | 685.17 | 1,850 | 7,204.95 | 29,843.67 | 0.67 | 1,170.7 | 5,672.6 | -17.26 | 20.79 | 99.67 | 4,114 | 7,722 | 43,626.33 | 24.6 | 1,501.47 | 814.53 | 47,738 | -19.13 | 6,278.33 | 135.8 | 11,815.87 | 4 | 21,438.64 | 3,533.67 | 1,179.67 | |
| TRPOevaluation_window=last 100 episodes, frame_count=40M, trials=32018.07 | 1,110.4 | 200.56 | 1,363.46 | 2,651.33 | 2,205.7 | 1,419,104.67 | 1,125.17 | 15,123.33 | 5,073.75 | 31.24 | 50.07 | 40.65 | 3,353.14 | 2,286.67 | 87,522.33 | 21,525.08 | -10.04 | 365.95 | -69.64 | 20.89 | 291.77 | 938.27 | 495.17 | -4.61 | 901.67 | 1,214.67 | 4,881.65 | 26,808 | 0 | 1,133.57 | 5,604.1 | -43.6 | 19.63 | 99.33 | 3,781.58 | 6,773.67 | 24,061.33 | 24.18 | 926.4 | 634.07 | 33,442.67 | -18.4 | 5,701 | 136.21 | 10,949.53 | 0 | 25,095.2 | 3,103 | 4,796.67 |