def get_agent(name):
a = TpFinalPlayer(name)
b = RandomPlayer('PC')
game = TpFinalGame([a, b])
print ('Training with a random player, please wait...')
game.agents = [a, b]
per = PerformanceCounter(game.agents, ['QLearnerA', 'QLearnerB'])
for i in range(3000):
game.run()
per.show_statistics()
return a
else:
return -1
return 0
if __name__ == '__main__':
a = TicTacToePlayer('X')
b = RandomPlayer('O')
c = HumanPlayer('O')
game = TicTacToeGame([a, b])
print ('Training with a random player, please wait...')
game.agents = [a, b]
for i in range(3000):
game.run()
a.dump('qlearner_agent')
d = TicTacToePlayer.load('qlearner_agent')
d.play_with = 'O'
game.agents = [a, d]
per = PerformanceCounter(game.agents, ['QLearnerA', 'QLearnerD'])
for i in range(3000):
game.run()
per.show_statistics()
game.agents = [a, c]
print ('Do you like to play?')
game.run()
print game.state
self.last_agent = None
random.shuffle(self.agents)
self.run()
def reward(self, state, agent):
if self.is_completed(state):
if self.last_agent is agent:
return -1
else:
return 1
return 0
def get_agent(name):
res = Player(name)
rand = RandomPlayer('random')
env = TrainingEnvironment([res, rand])
for i in range(3000):
env.play()
return res
if __name__ == '__main__':
res = Player('name')
rand = RandomPlayer('random')
env = TrainingEnvironment([res, rand])
per = PerformanceCounter(env.agents, ['q', 'random'])
for i in range(3000):
env.play()
per.show_statistics()
def is_completed(self, state):
return state in self.rewards.keys()
def reward(self, state, agent):
return self.rewards.get(state, -0.08)
class WumpusProblem(RLProblem):
def actions(self, state):
actions = ['up', 'down', 'left', 'rigth']
return actions
if __name__ == '__main__':
agent = TDQLearner(WumpusProblem(),
temperature_function=make_exponential_temperature(1000, 0.01),
discount_factor=0.8)
game = WumpusEnvironment(agent)
p = PerformanceCounter([agent], ['Q-learner Agent'])
print 'Training...'
for i in range(10000):
game.run()
p.show_statistics()
game.run(viewer=WumpusViewer(game))
def is_completed(self, state):
return state in self.rewards.keys()
def make_reward(self, state, agent):
return self.rewards.get(state, -0.08)
class GridProblem(RLProblem):
def actions(self, state):
actions = ['up', 'down', 'left', 'rigth']
return actions
if __name__ == '__main__':
agent = TDQLearner(GridProblem(),
temperature_function=make_exponential_temperature(1000, 0.005),
discount_factor=0.9)
game = GridGame(agent)
p = PerformanceCounter([agent], ['Q-learner Agent'])
for i in range(3000):
game.run()
p.show_statistics()
game.run(viewer=Viewer())
