Soft Actor-Critic for Discrete Action Settings

16 Oct 2019  ·  Petros Christodoulou ·

Soft Actor-Critic is a state-of-the-art reinforcement learning algorithm for continuous action settings that is not applicable to discrete action settings. Many important settings involve discrete actions, however, and so here we derive an alternative version of the Soft Actor-Critic algorithm that is applicable to discrete action settings. We then show that, even without any hyperparameter tuning, it is competitive with the tuned model-free state-of-the-art on a selection of games from the Atari suite.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Atari Games Atari 2600 Alien SAC Score 216.9 # 47
Atari Games Atari 2600 Amidar SAC Score 7.9 # 48
Atari Games Atari 2600 Assault SAC Score 350 # 45
Atari Games Atari 2600 Asterix SAC Score 272 # 49
Atari Games Atari 2600 Battle Zone SAC Score 4386.7 # 46
Atari Games Atari 2600 Beam Rider SAC Score 432.1 # 48
Atari Games Atari 2600 Breakout SAC Score 0.7 # 58
Atari Games Atari 2600 Crazy Climber SAC Score 3668.7 # 48
Atari Games Atari 2600 Enduro SAC Score 0.8 # 44
Atari Games Atari 2600 Freeway SAC Score 4.4 # 51
Atari Games Atari 2600 Frostbite SAC Score 59.4 # 53
Atari Games Atari 2600 James Bond SAC Score 68.3 # 44
Atari Games Atari 2600 Kangaroo SAC Score 29.3 # 46
Atari Games Atari 2600 Ms. Pacman SAC Score 690.9 # 44
Atari Games Atari 2600 Pong SAC Score -20.98 # 50
Atari Games Atari 2600 Q*Bert SAC Score 280.5 # 53
Atari Games Atari 2600 Road Runner SAC Score 305.3 # 42
Atari Games Atari 2600 Seaquest SAC Score 211.6 # 55
Atari Games Atari 2600 Space Invaders SAC Score 160.8 # 53
Atari Games Atari 2600 Up and Down SAC Score 250.7 # 44

Methods


No methods listed for this paper. Add relevant methods here