def play(agent):
start_data = api_game_start()
session_id = start_data['session_id']
grid = start_data['grid']
turn = 0
result = []
while (1):
# agent にはどんな形で渡せばよい?
# lib/pybrain/rl/experiments/experiment.py
# lib/pybrain/rl/environments/mazes/tasks/mdp.py
agent.integrateObservation(numpy.array(grid).ravel())
move = agent.getAction()
data = api_move(session_id, move[0])
if grid == data['grid']:
agent.giveReward(numpy.array([-50]))
rand_move = numpy.random.randint(4, size=1)[0]
data = api_move(session_id, rand_move)
else:
agent.giveReward(numpy.array([data['points']]))
turn += 1
grid = data['grid']
print_state(turn, data)
if data['over']:
print
#print data
break
return data['score']
def play(agent):
start_data = api_game_start()
session_id = start_data['session_id']
grid = start_data['grid']
turn = 0
result = []
while(1):
# agent にはどんな形で渡せばよい?
# lib/pybrain/rl/experiments/experiment.py
# lib/pybrain/rl/environments/mazes/tasks/mdp.py
agent.integrateObservation(numpy.array(grid).ravel())
move = agent.getAction()
data = api_move(session_id, move[0])
if grid == data['grid']:
agent.giveReward(numpy.array([-50]))
rand_move = numpy.random.randint(4, size=1)[0]
data = api_move(session_id, rand_move)
else:
agent.giveReward(numpy.array([data['points']]))
turn += 1
grid = data['grid']
print_state(turn, data)
if data['over']:
print
#print data
break
return data['score']
def play(agent):
#start_data = api_game_start()
start_data = api_simple_game_start()
session_id = start_data['session_id']
grid = start_data['grid']
next_point = 100
turn = 0
result = []
while(1):
# agent にはどんな形で渡せばよい?
# lib/pybrain/rl/experiments/experiment.py
# lib/pybrain/rl/environments/mazes/tasks/mdp.py
agent.integrateObservation(numpy.array(grid).ravel())
move = agent.getAction()
#data = api_move(session_id, move[0])
data = api_move(session_id, numpy.random.randint(4, size=1)[0])
if grid == data['grid']:
agent.giveReward(numpy.array([-50]))
data = api_move(session_id, numpy.random.randint(4, size=1)[0])
else:
agent.giveReward(numpy.array([data['points']]))
# # 100 point 毎に報酬を付与するやり方.
# if next_point < data['score']:
# print 'get point'
# agent.giveReward(numpy.array([data['score']]))
# next_point += 100
# else:
# agent.giveReward(numpy.array([0]))
turn += 1
grid = data['grid']
print_state(turn, data)
if data['over']:
print
#print data
break
return data['score']