List your company

Reinforcement Learning on Atari 2600 Games Breakdown

1,399,753Avg Reward (baseline)PPO with RUDDER
-86,671.16299,227.42685,1261,071,024.58Jun 20, 2018

Evaluation results

MethodLinks
PPO with RUDDERGame=Atlantis, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.061,399,7531,655,46418.33,568,5133,641,5832
PPO with RUDDERGame=CrazyClimber, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.06125,249106,076-15.3140,080132,480-5.4
PPO with RUDDERGame=KungFuMaster, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0649,37751,9845.366,88378,46017.3
PPO with RUDDERGame=YarsRevenge, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0646,70760,57729.7109,083178,43863.6
PPO with RUDDERGame=DemonAttack, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0628,68446,11960.8464,151400,370-13.7
PPO with RUDDERGame=Centipede, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0625,16223,029-8.553,00036,383-31.4
PPO with RUDDERGame=Hero, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0622,74612,162-46.532,38315,068-53.5
PPO with RUDDERGame=Asteroids, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0622,0654,905-77.8310,505154,479-50.2
PPO with RUDDERGame=RoadRunner, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0620,31427,99237.835,35336,7173.9
PPO with RUDDERGame=VideoPinball, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0619,73819,196-2.715,24816,83610.4
PPO with RUDDERGame=Phoenix, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0615,01339,24716146,758190,123307
PPO with RUDDERGame=BattleZone, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0612,87017,02332.324,66728,06713.8
PPO with RUDDERGame=Qbert, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0612,5228,290-33.828,76316,631-42.2
PPO with RUDDERGame=Asterix, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0610,55418,05471.129,513102,930249
PPO with RUDDERGame=Krull, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.069,0298,027-11.110,3688,202-20.9
PPO with RUDDERGame=NameThisGame, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.068,39010,54525.710,96217,24257.3
PPO with RUDDERGame=Gopher, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.068,1264,090-49.7102,91623,367-77.3
PPO with RUDDERGame=Zaxxon, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.066,9007,4988.712,12010,613-12.4
PPO with RUDDERGame=StarGunner, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.066,34629,01635718,38062,593241
PPO with RUDDERGame=ChopperCommand, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.066,1835,244-15.210,8179,573-11.5
PPO with RUDDERGame=Assault, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.065,7884,242-26.710,37311,3059
PPO with RUDDERGame=MsPacman, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.064,0965,00522.26,4466,9848.3
PPO with RUDDERGame=WizardOfWor, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.063,8613,024-21.76,4805,950-8.2
PPO with RUDDERGame=TimePilot, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.063,7904,208114,5335,56322.7
PPO with RUDDERGame=Frostbite, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.063,1724,77050.45,0927,43946.1
PPO with RUDDERGame=Kangaroo, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.062,7559,7642545,36013,500152
PPO with RUDDERGame=BeamRider, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.062,3724,50689.93,9946,74268.8
PPO with RUDDERGame=Alien, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.061,8783,08764.43,2185,70377.3
PPO with RUDDERGame=Solaris, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.061,3931,91837.76161,827197
PPO with RUDDERGame=Berzerk, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.061,2611,3416.41,9302,0928.4
PPO with RUDDERGame=Gravitar, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.061,2041,41517.51,8382,23321.5
PPO with RUDDERGame=Seaquest, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.061,1052,4621231,6164,770195
PPO with RUDDERGame=BankHeist, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.069361,19427.51,0781,33523.8
PPO with RUDDERGame=Amidar, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.06787724-81,2421,054-15.1
PPO with RUDDERGame=SpaceInvaders, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.067781,10642.11,2811,86045.2
PPO with RUDDERGame=Enduro, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.067597772.52,2011,339-39.2
PPO with RUDDERGame=Venture, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0673888520.18201,35064.6
PPO with RUDDERGame=Breakout, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.06217153-29.5430352-18.1
PPO with RUDDERGame=Tutankham, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0612315122.714016316.3
PPO with RUDDERGame=PrivateEye, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0610234.1-66.410033.3-66.7
PPO with RUDDERGame=Boxing, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.069894.7-3.410099.5-0.5
PPO with RUDDERGame=Bowling, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0661.517919156.3192241
PPO with RUDDERGame=Freeway, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0626.725.4-4.83231.4-1.9
PPO with RUDDERGame=Robotank, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0624.932.731.332.247.346.9
PPO with RUDDERGame=FishingDerby, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0619.511.7-39.95236.3-30.3
PPO with RUDDERGame=Pong, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.0619.218.5-3.921210
PPO with RUDDERGame=MontezumaRevenge, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.060038.4000
PPO with RUDDERGame=IceHockey, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.06-3.1-1.939.4-1.41171
PPO with RUDDERGame=Pitfall, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.06-8.4-5.534-750100
PPO with RUDDERGame=DoubleDunk, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.06-9.2-13.1-41.7-0.3-5.1-1,825
PPO with RUDDERGame=Tennis, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.06-13.5-13.50.2-4-5.3-32.8
PPO with RUDDERGame=Skiing, Total frames=200M, Starting condition=no-op, Max episode frames=108K, Safe exploration=false2018.06-29,501-29,911-1.4-29,977-29,9780