def make_move(self, board, cardDeck, gameEnd):
self.state = board.getState() + cardDeck.getState()
if gameEnd:
new_state = board.getState() + cardDeck.getState()
return new_state
p = random.uniform(0, 1)
if p < self.exp_factor:
new_state = self.make_optimal_move(board, cardDeck)
else:
moves = board.getValidMoves()
idx = random.choice(moves)
board.makeMove(idx, cardDeck.currentCard)
new_state = board.getState() + cardDeck.getState()
return new_state
def learn_state(self, board, cardDeck, gameEnd):
state = board.getState() + cardDeck.getState()
target = self.calc_target(board, cardDeck, gameEnd)
self.train_model(target, 10)
self.prev_state = state
def make_optimal_move(self, board, cardDeck):
moves = board.getValidMoves()
if len(moves) == 1:
board.makeMove(moves[0], cardDeck)
return board.getState() + cardDeck.getState()
temp_move_list = []
v = -float('Inf')
for idx in moves:
v_temp = []
temp_state = board.previewMove(idx, cardDeck.currentCard)
simCardDraws = 50
while simCardDraws > 0:
simCardDraws -= 1
temp_state_op = temp_state + cardDeck.previewNextCard()
v_temp.append(self.calc_value(temp_state_op))
# deletes Nones
v_temp = list(filter(None.__ne__, v_temp))
if len(v_temp) != 0:
v_temp = np.average(v_temp)
else:
# encourage exploration
v_temp = 50
if v_temp > v:
temp_move_list = [idx]
v = v_temp
elif v_temp == v:
temp_move_list.append(idx)
try:
board.makeMove(random.choice(temp_move_list), cardDeck)
except ValueError:
print('temp state:', temp_state_list)
raise Exception('temp state empty')
return board.getState() + cardDeck.getState()
def calc_target(self, board, cardDeck, gameEnd):
v_s = self.calc_value(self.prev_state)
R = self.reward(board)
if gameEnd:
v_s_tag = 0
else:
v_s_tag = self.calc_value(board.getState() + cardDeck.getState())
target = np.array(v_s + self.alpha * (R + v_s_tag - v_s))
return target
def make_move(self, board, cardDeck, gameEnd):
idx = int(input('Choose station number: '))
board.makeMove(idx, cardDeck.currentCard)
return board.getState() + cardDeck.getState()