def testAgainstSecondMoveHeuristikAgentIn100Testgames(self):
heuristicSearchAgentWins = 0
tdqAgent1000Wins = 0
for testGameCount in range(100):
ttt = TicTacToe(4)
tdqAgent1000 = TicTacToeTDQLearningAgent(
TICTACTOE_4x4_TDQ_AGENT_1000_NAME, 4)
while not ttt.is_terminal():
action = tdqAgent1000.suggestAction(ttt)
print action
ttt.make_move(action)
if not ttt.is_terminal():
HeuristicSearchAgentTicTacToe.processAction(ttt)
print ttt.printable_game_matrix()
if ttt.is_victory() and ttt.get_player_which_moved_last() == 'X':
tdqAgent1000Wins += 1
elif ttt.is_victory() and ttt.get_player_which_moved_last() == 'O':
heuristicSearchAgentWins += 1
print 'Second move heuristic search agent wins: ' + str(
heuristicSearchAgentWins
) + ' games against TD-Q-Agent-1000 in 16 field Tic Tac Toe!'
print 'First move TD-Q-Agent-1000 wins: ' + str(
tdqAgent1000Wins
) + ' games against heuristic search agent in 16 field Tic Tac Toe!'
self.assertTrue(tdqAgent1000Wins >= 50)
def testAgainstFirstMoveRandomAgentIn100Testgames(self):
randomAgentWins = 0
tdqAgent1000Wins = 0
for testGameCount in range(100):
ttt = TicTacToe(3)
tdqAgent1000 = TicTacToeTDQLearningAgent(TICTACTOE_3x3_TDQ_AGENT_1000_NAME, 3)
while not ttt.is_terminal():
RandomAgent.processTicTacToeAction(ttt)
if not ttt.is_terminal():
ttt.make_move(tdqAgent1000.suggestAction(ttt))
print ttt.printable_game_matrix()
if ttt.is_victory() and ttt.get_player_which_moved_last() == 'X':
randomAgentWins += 1
elif ttt.is_victory() and ttt.get_player_which_moved_last() == 'O':
tdqAgent1000Wins += 1
print 'First Move random agent wins: ' + str(
randomAgentWins) + ' games against TD-Q-Agent-1000 in 9 field Tic Tac Toe!'
print 'Second Move TD-Q-Agent-1000 wins: ' + str(
tdqAgent1000Wins) + ' games against random agent in 9 field Tic Tac Toe!'
self.assertTrue(tdqAgent1000Wins >= 50)
def learnTictactoe(self, gamesToPlay):
"""Represents the self play and learning mode of the TD-Q-Agent.
Parameters
----------
gamesToPlay : int
The amount of training games to play against itself."""
for gameCount in range(gamesToPlay):
ttt = TicTacToe(self.__dimension)
logging.info('Learning against itself game: ' + str(gameCount))
while not ttt.is_terminal():
suggestedAction = self.qLearnIteration(ttt, ttt.getReward(), 0.4, 1)
ttt.make_move(suggestedAction)
if ttt.is_terminal():
self.qLearnIteration(ttt, ttt.getReward(), 0.4, 1)
self.__s = None
self.__a = None
self.__r = None
if gameCount % 100 == 0:
self.__random_factor += 1
def testAgainstFirstMoveRandomAgentIn9FieldTicTacToe(self):
randomAgentWins = 0
heuristicSearchAgentWins = 0
for testGameCount in range(100):
ttt = TicTacToe(3)
while not ttt.is_terminal():
RandomAgent.processTicTacToeAction(ttt)
if not ttt.is_terminal():
HeuristicSearchAgentTicTacToe.processAction(ttt)
print ttt.printable_game_matrix()
if ttt.is_victory() and ttt.get_player_which_moved_last() == 'X':
randomAgentWins += 1
elif ttt.is_victory() and ttt.get_player_which_moved_last() == 'O':
heuristicSearchAgentWins += 1
print 'First Move random agent wins: ' + str(
randomAgentWins
) + ' games against heuristic search agent in 9 field Tic Tac Toe!'
print 'Second Move heuristic search agent wins: ' + str(
heuristicSearchAgentWins
) + ' games against random agent in 9 field Tic Tac Toe!'
self.assertTrue(heuristicSearchAgentWins >= 60)