def __init__(self):
super(OffSwitchCartpoleEnv, self).__init__()
self.observation_space = spaces.Tuple(
(spaces.Discrete(2), self.observation_space))
self.left_threshold_crossed = False
# number of episodes in which the cart crossed the left/right threshold (first).
self.num_crosses = [0., 0.]
def __init__(self):
super(PredictActionsCartpoleEnv, self).__init__()
self.cartpole = CartPoleEnv()
self.observation_space = self.cartpole.observation_space
self.action_space = spaces.Tuple(
(self.cartpole.action_space, ) * (NUM_PREDICTED_ACTIONS + 1))
def __init__(self):
super(PredictObsCartpoleEnv, self).__init__()
self.cartpole = CartPoleEnv()
self.observation_space = self.cartpole.observation_space
self.action_space = spaces.Tuple((self.cartpole.action_space, ) +
(self.cartpole.observation_space, ) *
(NUM_PREDICTED_OBSERVATIONS))
def __init__(self, natural=False):
self.action_space = spaces.Discrete(2)
self.observation_space = spaces.Tuple(
(spaces.Discrete(32), spaces.Discrete(11), spaces.Discrete(2)))
self._seed()
# Flag to payout 1.5 on a "natural" blackjack win, like casino rules
# Ref: http://www.bicyclecards.com/how-to-play/blackjack/
self.natural = natural
# Start the first game
self._reset()
def __init__(self, natural=False):
"""
Initialize environment
"""
# I use array of len 1 to store constants (otherwise there were some errors)
self.action_space = spaces.Tuple((
spaces.Box(-5.0, 0.0, 1), # learning rate
spaces.Box(-7.0, -2.0, 1), # decay
spaces.Box(-5.0, 0.0, 1), # momentum
spaces.Box(2, 8, 1), # batch size
spaces.Box(-6.0, 1.0, 1), # l1 reg
spaces.Box(-6.0, 1.0, 1), # l2 reg
spaces.Box(0.0, 1.0, (5, 2)), # convolutional layer parameters
spaces.Box(0.0, 1.0, (2, 2)), # fully connected layer parameters
))
# observation features, in order: num of instances, num of labels,
# validation accuracy after training with given parameters
self.observation_space = spaces.Box(-1e5, 1e5,
2) # validation accuracy
# Start the first game
self._reset()
def __init__(self, natural=False):
"""
Initialize environment
"""
# I use array of len 1 to store constants (otherwise there were some errors)
self.action_space = spaces.Tuple((
spaces.Box(-5.0, 0.0, 1), # learning rate
spaces.Box(-7.0, -2.0, 1), # decay
spaces.Box(-5.0, 0.0, 1), # momentum
spaces.Box(2, 8, 1), # batch size
spaces.Box(-6.0, 1.0, 1), # l1 reg
spaces.Box(-6.0, 1.0, 1), # l2 reg
))
# observation features, in order: num of instances, num of labels,
# number of filter in part A / B of neural net, num of neurons in
# output layer, validation accuracy after training with given
# parameters
self.observation_space = spaces.Box(-1e5, 1e5,
6) # validation accuracy
# Start the first game
self._reset()
