def opponent(self):
if self.mark.opponent() == 1:
return Maru()
elif self.mark.opponent() == 0:
return Empty()
elif self.mark.opponent() == -1:
return Batsu()
def is_draw(self):
'''
盤面の状態が引き分けかどうか
return:
True:引き分け
False:引き分けではない（勝負がついていない）
'''
if len(self.get_valid_actions()) == 0 and not self.is_win(Mark(
Maru())) and not self.is_win(Mark(Batsu())):
return True
return False
def is_end(self):
'''
盤面の状態がゲーム修了かどうか
return:
True:修了
False:修了していない（勝負がついていない）
'''
if self.is_win(Mark(Maru())) or self.is_win(Mark(Batsu())) or len(
self.get_valid_actions()) == 0:
return True
return False
# 勝者の報酬
current_player.learn(1, True)
# 敗者の報酬
self.players[result.opponent().to_int()].learn(-1, True)
if verbose:
print("%s" % (state.to_array()))
print("-" * 5)
state.output()
print("-" * 5)
print("%s win!!!" %
(self.players[current_player_mark].mark.to_string()))
break
elif state.is_draw():
result = Mark(Empty())
for player in self.players.itervalues():
player.learn(0, True)
if verbose:
state.output()
print("draw.")
break
current_player_mark = self.players[
current_player_mark].mark.opponent().to_int()
#print("="*30)
if __name__ == '__main__':
player1 = Player(Mark(Maru()))
player2 = Player(Mark(Batsu()))
game = Game(player1, player2)
game.start(verbose=True)
print new_state.get(0).to_int()
print state1.to_array()
print new_state.to_array()
print state1.win_state
print state1.is_win(Mark(Maru()))
maru1 = Mark(Maru())
maru2 = Mark(Maru())
print maru1.to_int() == maru2.to_int()
new_state2 = new_state.set(1, Mark(Maru()))
new_state3 = new_state2.set(2, Mark(Maru()))
print new_state3.to_array()
print new_state3.is_win(Mark(Maru()))
print new_state2.is_draw()
print new_state2.is_end()
print new_state3.is_end()
new_state = state1.set(0, Mark(Batsu()))
new_state2 = new_state.set(1, Mark(Maru()))
new_state4 = new_state2.set(2, Mark(Batsu()))
new_state5 = new_state4.set(3, Mark(Maru()))
new_state6 = new_state5.set(4, Mark(Batsu()))
new_state7 = new_state6.set(5, Mark(Batsu()))
new_state8 = new_state7.set(6, Mark(Maru()))
new_state9 = new_state8.set(7, Mark(Batsu()))
new_state10 = new_state9.set(8, Mark(Maru()))
print new_state10.to_array()
print new_state10.is_draw()
new_state10.output()
state11 = State()
state12 = State()
状態
報酬
事後状態から行動を行った後の状態
'''
if next_state is None:
next_state_value = 0.0
else:
#状態価値テーブルから価値を取得
next_state_value = self.value[next_state]
# 行動価値の更新式
self.value[state] += self.step_size * (reward + next_state_value -
self.value[state])
#print("value[state] %f", self.value[state])
if __name__ == '__main__':
value = Value()
state = State()
#print value.get_value(state)
#print value.get_max_action(state, Mark(Maru()))
new_state = state.set(3, Mark(Maru()))
new_state1 = state.set(4, Mark(Maru()))
new_state2 = new_state.set(2, Mark(Batsu()))
value.update(state, 10, new_state)
value.update(new_state, 10, new_state2)
value.update(new_state, 10, new_state2)
value.update(new_state1, 10, new_state2)
value.update(new_state1, 100, new_state2)
#value.update(state, -3, new_state2)
print value.get_max_action(state, Mark(Maru()))
from maru_mark import Maru
from batsu_mark import Batsu
from empty_mark import Empty
from tic_tac_toe_state import State
from tic_tac_toe_game import Game
# 共通行動価値テーブル
value = Value()
print("Sarsa method:")
print("Training com1 and com2.")
com_1 = SarsaCom(Mark(Maru()), value)
com_2 = SarsaCom(Mark(Batsu()), value)
print("Input the number of iterations:")
iterations = 10000
while(True):
input_line = raw_input()
if input_line.isdigit():
iterations = int(input_line)
break
else:
print("Input number:")
for i in xrange(iterations):
game = Game(com_1, com_2)
game.start()
#encoding: utf-8
'''
tic tac toe のSarsa λ実行クラス
'''
from tic_tac_toe_sarsa_r_com import SarsaRCom
from tic_tac_toe_nn_com import SarsaNNCom
from mark import Mark
from maru_mark import Maru
from batsu_mark import Batsu
from tic_tac_toe_game import Game
import dill
com_1 = SarsaNNCom(Mark(Maru()), 0.1, 0.1, 0.6)
com_2 = SarsaRCom(Mark(Batsu()), 0.1, 0.1, 0.6)
iterations = 100000
print("Input the number of iterations (%d):" % (iterations))
while(True):
input_line = raw_input()
if input_line.isdigit():
iterations = int(input_line)
break
elif input_line == '':
break
else:
print("Input number:")
# 学習
for i in xrange(iterations):
# モデルのロード
with open('tic_tac_toe_com_1_sarsa_r.pkl', 'rb') as f:
com_1 = dill.load(f)
with open('tic_tac_toe_com_2_sarsa_r.pkl', 'rb') as f:
com_2 = dill.load(f)
while (True):
print("Select a type of fight [1, 2, 3, q]")
print("1: human vs com2")
print("2: com1 vs human")
print("3: com1 vs com2")
print("q: quit")
type_of_fight = 1
input_line = raw_input()
if input_line.isdigit():
type_of_fight = int(input_line)
else:
if input_line == 'q':
break
continue
if type_of_fight == 1:
game = Game(Player(Mark(Maru())), com_2)
elif type_of_fight == 2:
game = Game(com_1, Player(Mark(Batsu())))
elif type_of_fight == 3:
game = Game(com_1, com_2)
game.start(True)
def opponent(self):
if self.mark.opponent() == 1:
return Maru()
elif self.mark.opponent() == 0:
return Empty()
elif self.mark.opponent() == -1:
return Batsu()
def to_int(self):
return self.mark.to_int()
def to_string(self):
return self.mark.to_string()
if __name__ == '__main__':
maru = Mark(Maru())
print maru.to_string()
print maru.to_int()
print maru.opponent().to_string()
print '=' * 10
batsu = Mark(Batsu())
print batsu.to_string()
print batsu.to_int()
print batsu.opponent().to_string()
print '=' * 10
empty = Mark(Empty())
print empty.to_string()
print empty.to_int()
print empty.opponent().to_string()