damquangtuan / w_mcts_atari

To reproduce the result in Figure 1, run python alphazero.py --game=CartPole-v1 --algorithm=uct --epsilon=0.1 python alphazero.py --game=CartPole-v1 --algorithm=ments --epsilon=0.1 python alphazero.py --game=CartPole-v1 --algorithm=rents --epsilon=0.1 python alphazero.py --game=CartPole-v1 --algorithm=tsallis --epsilon=0.1

python alphazero.py --game=Acrobot-v1 --algorithm=uct --epsilon=0.1 python alphazero.py --game=Acrobot-v1 --algorithm=ments --epsilon=0.1 python alphazero.py --game=Acrobot-v1 --algorithm=rents --epsilon=0.1 python alphazero.py --game=Acrobot-v1 --algorithm=tsallis --epsilon=0.1

To reproduce results in Table 2: to run MENTS: python alphago.py --game=AsterixNoFrameskip-v4 --tau=0.02 --algorithm=ments --epsilon=0.

Replace algorithm and game parameters with the following: Please note that due to the maximum size allowed to submit, we have to remove 7 models from the model folder. List of games: AlienNoFrameskip-v4, AmidarNoFrameskip-v4, AsteroidsNoFrameskip-v4, BankHeistNoFrameskip-v4, BowlingNoFrameskip-v4, CentipedeNoFrameskip-v4, DemonAttackNoFrameskip-v4, GopherNoFrameskip-v4, PhoenixNoFrameskip-v4, RobotankNoFrameskip-v4, WizardOfWorNoFrameskip-v4, AtlantisNoFrameskip-v4, EnduroNoFrameskip-v4, FrostbiteNoFrameskip-v4, HeroNoFrameskip-v4, MsPacmanNoFrameskip-v4, SolarisNoFrameskip-v4, BreakoutNoFrameskip-v4, AsterixNoFrameskip-v4, BeamRiderNoFrameskip-v4, QbertNoFrameskip-v4, SeaquestNoFrameskip-v4, SpaceInvadersNoFrameskip-v4 List of algorithm: uct (parameter: c) power-uct (parameter: c and p) maxmcts(parameter: lambda_const) rents (parameter: tau and epsilon) ments (parameter: tau and epsilon)

w_mcts_atari