def mm_find_best():
"""MiniMax find best move (max_layers=3)."""
t = mm_tg()
t.vline(3, 0, 15, t.body_of(1))
mm = MiniMax(t, mm_player())
mm.find_best_move(max_layers=3)
def test_catch(tg, player):
"""Test catching the other player."""
tg.vline(3, 0, 15, tg.body_of(1))
mm = MiniMax(tg, player)
weight, move = mm.find_best_move(max_layers=3)
assert move == 'LEFT'
def __init__(self, shape, depth_lim=9):
Computer.__init__(self, shape)
evaluator = Evaluator()
self.mini_max_obj = MiniMax(evaluator.eval, self.shape,
self.other_shape())
self.depth_lim = depth_lim
self.name_str = 'MiniMax'
def action_called(self):
# get the button that called the action
button = self.sender()
# disable button
button.setEnabled(False)
# traverse through the buttons to get the coords of button (there is probably a better way to do this)
for i in range(len(self.push_list)):
for j in range(len(self.push_list)):
if button == self.push_list[i][j]:
move = (i, j)
break
if self.is_ai is True:
# set the text of the button to X
button.setText(self.game.get_X_player())
# make the move in the game
self.game.make_move(self.game.get_X_player(), move)
# if the game is unfinished, make the AI move
if self.game.get_status() == "UNFINISHED":
game_object = copy.deepcopy(self.game)
ai = MiniMax(game_object)
ai_move = ai.minimax(game_object)
button = self.push_list[ai_move[0]][ai_move[1]]
button.setText(self.game.get_O_player())
button.setEnabled(False)
self.game.make_move(self.game.get_O_player(), ai_move)
else:
turn = self.game.get_current_player()
if turn == self.game.get_X_player():
button.setText(self.game.get_X_player())
self.game.make_move(self.game.get_X_player(), move)
else:
button.setText(self.game.get_O_player())
self.game.make_move(self.game.get_O_player(), move)
# determine if there is a win or draw
win = self.game.get_status()
# set the game status label to empty text
text = ""
if win == "X_WON":
text = "X WON"
if win == "O_WON":
text = "O WON"
if win == "DRAW":
text = "DRAW"
self.label.setText(text)
def go(self):
"""Act depending if we see others."""
if self.can_see_others():
mm = MiniMax(self.grid, self.players[self.my_number],
full_bfs=self.full_bfs)
weight, move = mm.find_best_move(max_layers=self.max_layers,
max_layer_size=self.max_layer_size)
mm.unlink_states()
if weight > 0:
return move
return self.go_wander()
def aiPlay(self):
m = MiniMax(self.gameBoard)
best_move = m.bestMove(5, self.gameBoard, self.currentTurn)
#print("best move: ", best_move)
self.playPiece(best_move)
# print('\n\nmove %d: Player %d, column %d\n' % (self.pieceCount, self.currentTurn, randColumn+1))
if self.currentTurn == 1:
self.currentTurn = 2
elif self.currentTurn == 2:
self.currentTurn = 1
def test_obvious_win(self):
#player 1 should go for the win here
gameArr = [
[0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0],
[1, 2, 0, 0, 0, 0, 0],
[1, 2, 0, 0, 0, 0, 0],
[1, 2, 0, 0, 0, 0, 0]
]
m = MiniMax(gameArr)
best_move = m.bestMove(5, gameArr, 1)
self.assertEqual(best_move, 0)
def test_diagnol_win(self):
# player 2 has win on diagnol
gameArr = [
[0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0],
[0, 1, 2, 1, 0, 0, 0],
[0, 1, 1, 2, 0, 0, 0],
[0, 2, 1, 2, 2, 0, 0],
[1, 2, 2, 2, 1, 0, 0]
]
m = MiniMax(gameArr)
best_move = m.bestMove(5, gameArr, 2)
self.assertEqual(best_move, 1)
def test_game_end_full(self):
# player 2 has win on diagnol
gameArr = [
[1, 2, 1, 2, 1, 2, 1],
[2, 1, 2, 1, 2, 1, 2],
[1, 1, 2, 2, 1, 1, 2],
[2, 2, 1, 1, 1, 2, 2],
[2, 1, 1, 2, 2, 1, 1],
[1, 2, 2, 2, 2, 2, 2]
]
m = MiniMax(gameArr)
best_move = m.bestMove(5, gameArr, 2)
self.assertEqual(best_move, None)
def test_game_end_win(self):
gameArr = [
[0, 0, 0, 0, 0, 0, 0],
[0, 2, 0, 0, 0, 0, 0],
[0, 1, 2, 1, 0, 0, 0],
[0, 1, 1, 2, 0, 0, 0],
[0, 2, 1, 2, 2, 0, 0],
[1, 2, 2, 2, 1, 0, 0]
]
m = MiniMax(gameArr)
best_move = m.bestMove(5, gameArr, 2)
self.assertEqual(best_move, None)
def test_block(self):
#palyer 2 should block player 1 here
gameArr = [
[0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 1, 0],
[0, 2, 0, 0, 0, 1, 0],
[0, 2, 2, 1, 0, 1, 0]
]
m = MiniMax(gameArr)
best_move = m.bestMove(5, gameArr, 2)
self.assertEqual(best_move, 5)
def test_init(tg, player):
"""Test initialization."""
mm = MiniMax(tg, player)
assert mm.my_number == 1
assert mm.my_pos == tg.coords2index(2, 15)
assert mm.opponents == {2: tg.coords2index(0, 5)}
def fit(self):
random = RandomMove()
minimax = MiniMax(max_depth=9)
agents = np.array([random, self])
state = np.zeros(n_size * n_size)
for i in range(20001):
np.random.shuffle(agents)
extended_boards, extended_actions, rewards, unfinished_flags, _ = play(
agents)
for board_sequence, action_sequence in zip(extended_boards,
extended_actions):
for state, next_state, action, reward, unfinished in zip(
board_sequence[:-1], board_sequence[1:],
action_sequence, rewards, unfinished_flags):
state_hash = self.hash(state)
next_hash = self.hash(next_state)
self.q[state_hash][action] += self.alpha * (
reward +
self.gamma * unfinished * np.amax(self.q[next_hash]) -
self.q[state_hash][action])
if i % 1000 == 0:
print(f'iteration {i}\t\t\twin/draw/lose')
print('minimax vs. q learning', test([minimax, self]))
print('q learning vs. minimax', test([self, minimax]))
print('random vs. q learning', test([random, self]))
print('q learning vs. random', test([self, random]))
def test_first_layer(tg, player):
"""Test computing the first layer in open field."""
mm = MiniMax(tg, player)
mm.compute_next_layer()
assert len(mm.layers) == 1
assert len(mm.layers[0]) == 4
assert {state.moves[0].direction for state in mm.layers[0]} ==\
set(tg.DIRECTIONS.keys())
for state in mm.layers[0]:
assert len(state.moves) == 1
move = state.moves[0]
assert move.player_number == 1
assert move.is_mine is True
assert mm.my_pos + tg.DIRECTIONS[move.direction] == state.player2pos[1]
def get_agent(self, task_id, board, depth):
if task_id == self.GBFS:
return GBFS(board)
elif task_id == self.MINIMAX:
return MiniMax(board, depth)
else:
return AlphaBeta(board, depth)
def main():
minimax = MiniMax(max_depth=9)
mcts = MCTS()
random = RandomMove()
test([mcts, mcts])
print('\t\t\t\twin/draw/lose')
print('mcts vs. mcts', test([mcts, mcts]))
print('random vs. mcts', test([random, mcts]))
print('mcts vs. random', test([mcts, random]))
print('minimax vs. mcts', test([minimax, mcts]))
print('mcts vs. minimax', test([mcts, minimax]))
def test_second_layer(tg, player):
"""Test computing two layers in open field with opponent next to wall."""
mm = MiniMax(tg, player)
mm.compute_next_layer()
mm.compute_next_layer()
assert len(mm.layers) == 2
assert len(mm.layers[0]) == 4
assert len(mm.layers[1]) == 12
for state in mm.layers[0]:
assert state.next_player == 2
assert len(state.next_states) == 3
assert {nstate.moves[1].direction for nstate in state.next_states} ==\
set(tg.DIRECTIONS.keys()) - {'LEFT'}
for nstate in state.next_states:
assert nstate.player_number == 2
assert nstate.prev_state == state
assert len(nstate.moves) == 2
move = nstate.moves[1]
assert move.is_mine == False
assert move.player_number == 2
assert mm.opponents[2] + tg.DIRECTIONS[move.direction] ==\
nstate.player2pos[2]
class PrunePlayer(Computer):
def __init__(self, shape, depth_lim=9):
Computer.__init__(self, shape)
evaluator = Evaluator()
self.mini_max_obj = MiniMax(evaluator.eval, self.shape,
self.other_shape())
self.depth_lim = depth_lim
self.name_str = 'Prune'
def get_move(self):
if self.current_board.count_empty() == 9:
pos = [random.choice([0, 1, 2]), random.choice([0, 1, 2])]
else:
start_time = time.time()
score, pos = self.mini_max_obj.minimax_alphabeta(
self.current_board, self.depth_lim, self.shape,
self.other_shape())
end_time = time.time()
print(f'Elapsed time (Pruned): {end_time - start_time}')
return pos
class SmartestPlayer(Computer):
def __init__(self, shape, depth_lim=9):
Computer.__init__(self, shape)
evaluator = Evaluator()
self.mini_max_obj = MiniMax(evaluator.eval, self.shape,
self.other_shape())
self.depth_lim = depth_lim
self.name_str = 'MiniMax'
def get_move(self):
if self.current_board.count_empty() == 9:
pos = [random.choice([0, 1, 2]), random.choice([0, 1, 2])]
else:
start_time = time.time()
move = self.mini_max_obj.minimax(self.current_board,
self.depth_lim, self.shape,
self.other_shape())
end_time = time.time()
print(f'Elapsed time (Alpha-Beta): {end_time - start_time}')
pos = [move[1], move[2]]
return pos
def fit(self):
random = RandomMove()
minimax = MiniMax(max_depth=9)
agents = [minimax, self]
while self.states.shape[0] < self.training_size:
# np.random.shuffle(agents)
play(agents, self)
for iteration in range(self.n_episodes):
self.eps *= self.eps_decay
# np.random.shuffle(agents)
play(agents, self)
print('iteration:', iteration, 'eps:', self.eps)
for i in range(10):
self.replay()
if iteration % 10 == 0:
self.target_net.copy_weights(self.policy_net)
temp_eps = self.eps
self.eps = 0
print('\t\t\t\twin/draw/lose')
print('minimax vs. dqn', test([minimax, self]))
print('dqn vs. minimax', test([self, minimax]))
print('random vs. dqn', test([random, self]))
print('dqn vs. random', test([self, random]))
self.eps = temp_eps
from minimax import MiniMax
from alphabeta import AlphaBeta
from qlearning import QLearning
import time
game = TicTacToe()
user = int(input("Player1(X) or Player2(O):"))
ai = 2 if user == 1 else 1
ai_algorithm = input("""Choose your opponent:
1. MiniMax Algorithm
2. MiniMax with Alpha-Beta Pruning
3. Q-Learning Agent
""")
if ai_algorithm == "1":
agent = MiniMax(player=ai)
elif ai_algorithm == "2":
agent = AlphaBeta(player=ai)
elif ai_algorithm == "3":
agent = QLearning(player=ai)
agent.epsilon = 0
agent.load_q_table()
while True:
game.render()
print("-----------------------------------------------------------------")
if game.turn == user:
action = int(input("Action (0-8):"))
done = game.step(action)
if done:
game.render()
def minimax(self):
agent = MiniMax(self.board, self.depth)
board = agent.get_next_board()
agent.output_next_state(board)
agent.output_log()
def _evalState(self, s):
m = MiniMax(depth=self._depth)
if self._useDelta:
return min(max(m.getBoardScoreDelta(s), -12.0), 12.0)
else:
return min(max(m.getBoardScore(s), -12.0), 12.0)
# %% import numpy as np import matplotlib.pyplot as plt # %% from tictactoe import Board, X, O from minimax import MiniMax from expectiminimax import ExpectiMiniMax from plot import plot_board_score, clean_square # %% [markdown] heading_collapsed=true # ## Optimal adversary # %% hidden=true engine = MiniMax() b = Board() engine.search(b) # %% hidden=true engine[b] # %% hidden=true # %% hidden=true engine[Board((1, 0, 0, 0, 0, 0, 0, 0, 0))] # %% hidden=true a = (1, 0, 2, 0, 0) b = (0, 0, 2, 0, 0) tuple(a_ ^ b_ for a_, b_ in zip(a, b))