def user_control(self):
while True:
self.action = input("Enter a command or a number to train: ")
# Test classification percentage using both the test set and training set
if self.action[0] == "t":
self.test_percentage_training_and_test_set()
# Play forever
elif self.action[0] == "s":
self.results_length = float('inf')
return
# Play 50 games, using the neural net p or a random player r
elif self.action[0] == "p" or self.action[0] == "r":
self.results_length = 50
return
# Run the grading function
elif self.action[0] == "c":
if len(self.results_from_nn_playing) < 50:
self.results_length = 50
return
else:
self.print_comparison()
elif self.action[0] == "l":
self.collect_cases = not self.collect_cases
if self.collect_cases:
self.neural_network_cases = load_cases()
self.expectimax = Expectimax()
else:
self.errors = self.move_classifier.do_training(
epochs=int(self.action), errors=self.errors)
self.test_percentage_training_and_test_set()
print("Total time elapsed: " +
str(round((time() - self.start_time) / 60, 1)) + " min")
def __init__(self):
Frame.__init__(self)
self.grid()
self.master.title('Ganesh-2048')
self.grid_cells = []
self.initialise_grid()
self.initilaise_matrix()
self.update_grid_cells()
self.AI = Expectimax()
self.start_game()
self.mainloop()
def start_game(self):
print "avg", sum(self.results)/float(len(self.results) + 0.001)
if len(self.results) < 30:
self.game_board = Game2048(board=[[0,0,0,0],[0,0,0,0],[0,0,0,0],[0,0,0,0]])
self.board = self.game_board.board
self.game_board.generate_new_node()
self.depth = 4
self.move_count = 0
self.expectimax = Expectimax()
self.draw_board()
self.time = time()
self.run_algorithm()
else:
print self.results
print "avg", sum(self.results)/float(len(self.results))
def get_strategy(strategy_type):
if strategy_type == "minimax":
return Minimax()
elif strategy_type == "expectimax":
return Expectimax()
elif strategy_type == "simple":
return Simple()
else:
return Random()
def __init__(self,
collect_cases=False,
use_merge_input_nodes=False,
depth=3):
self.collect_cases = collect_cases
self.use_merge_input_nodes = use_merge_input_nodes
self.depth = depth
if collect_cases:
self.neural_network_cases = load_cases()
self.expectimax = Expectimax()
self.results_from_nn_playing = []
self.results_from_random_playing = []
self.results = []
self.results_from_random_playing = [112] * 50
self.start_time = time()
self.print_commands()
self.setup_network()
self.user_control()
self.start_game()
class Game(Frame):
def __init__(self):
Frame.__init__(self)
self.grid()
self.master.title('Ganesh-2048')
self.grid_cells = []
self.initialise_grid()
self.initilaise_matrix()
self.update_grid_cells()
self.AI = Expectimax()
self.start_game()
self.mainloop()
def start_game(self):
while True:
self.board.move(self.AI.retrievemove(self.board))
self.update_grid_cells()
self.add_random_number()
self.update_grid_cells()
if len(self.board.get_available_moves()) == 0:
self.game_over_display()
break
self.update()
def game_over_display(self):
for i in range(4):
for j in range(4):
self.grid_cells[i][j].configure(text="",
bg=BACKGROUND_COLOR_CELL_EMPTY)
self.grid_cells[0][0].configure(text="best",
bg=BACKGROUND_COLOR_CELL_EMPTY)
top_4 = list(
map(int, reversed(sorted(list(self.board.grid.flatten())))))
self.grid_cells[0][1].configure(text=str(top_4[0]),
bg=BACKGROUND_COLOR_DICT[2048],
fg=CELL_COLOR_DICT[2048])
self.update()
def initialise_grid(self):
background = Frame(self,
bg=BACKGROUND_COLOR_GAME,
width=SIZE,
height=SIZE)
background.grid()
for i in range(GRID_LEN):
grid_row = []
for j in range(GRID_LEN):
cell = Frame(background,
bg=BACKGROUND_COLOR_CELL_EMPTY,
width=SIZE / GRID_LEN,
height=SIZE / GRID_LEN)
cell.grid(row=i,
column=j,
padx=GRID_PADDING,
pady=GRID_PADDING)
# font = Font(size=FONT_SIZE, family=FONT_FAMILY, weight=FONT_WEIGHT)
t = Label(master=cell,
text="",
bg=BACKGROUND_COLOR_CELL_EMPTY,
justify=CENTER,
font=FONT,
width=4,
height=2)
t.grid()
grid_row.append(t)
self.grid_cells.append(grid_row)
def gen(self):
return randint(0, GRID_LEN - 1)
def initilaise_matrix(self):
self.board = BoardDesign()
self.add_random_number()
self.add_random_number()
def update_grid_cells(self):
for i in range(GRID_LEN):
for j in range(GRID_LEN):
new_number = int(self.board.grid[i][j])
if new_number == 0:
self.grid_cells[i][j].configure(
text="", bg=BACKGROUND_COLOR_CELL_EMPTY)
else:
n = new_number
if new_number > 2048:
c = 2048
else:
c = new_number
self.grid_cells[i][j].configure(
text=str(n),
bg=BACKGROUND_COLOR_DICT[c],
fg=CELL_COLOR_DICT[c])
self.update_idletasks()
def add_random_number(self):
if randint(0, 99) < 90: #90% of times we will get 2
value = 2
else: #10 % of times we will get 4
value = 4
cells = self.board.get_available_cells()
pos = cells[randint(0, len(cells) - 1)] if cells else None
if pos is None:
return None
else:
self.board.insert_number(pos, value)
return pos
class GameController():
NR_OF_TRAINING_CASES = 7000
NR_OF_TEST_CASES = 833
NR_OF_OUTPUT_NODES = 4
def __init__(self,
collect_cases=False,
use_merge_input_nodes=False,
depth=3):
self.collect_cases = collect_cases
self.use_merge_input_nodes = use_merge_input_nodes
self.depth = depth
if collect_cases:
self.neural_network_cases = load_cases()
self.expectimax = Expectimax()
self.results_from_nn_playing = []
self.results_from_random_playing = []
self.results = []
self.results_from_random_playing = [112] * 50
self.start_time = time()
self.print_commands()
self.setup_network()
self.user_control()
self.start_game()
def start_game(self):
if len(self.results) < self.results_length:
print("run nr", len(self.results))
self.game_board = Game2048(
board=[[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]])
self.board = self.game_board.board
self.game_board.generate_new_node()
self.move_count = 0
#self.draw_board()
self.time = time()
self.run_algorithm()
else:
print(self.results)
print("Largest tile", max(self.results))
print("Average tile", sum(self.results) / float(len(self.results)))
if self.action[0] == "p":
self.results_from_nn_playing = copy.copy(self.results)
elif self.action[0] == "r":
self.results_from_random_playing = copy.copy(self.results)
elif self.action[0] == "c":
self.results_from_nn_playing = copy.copy(self.results)
self.print_comparison()
self.results = []
self.user_control()
self.start_game()
def setup_network(self):
if self.use_merge_input_nodes:
number_of_input_nodes = 24
else:
number_of_input_nodes = 16
use_default = input("Use default [y/n]: ")
if use_default[0] == "y":
nodes_in_each_layer = [700]
activation_functions = [3, 4]
learning_rate = 0.02
bulk_size = 1
else:
nodes_in_each_layer = list(
map(
int,
input("Hidden nodes in each layer: ").replace(
" ", "").split(",")))
print("TanH: 1, Sigmoid: 2, Rectify: 3, Softmax: 4")
activation_functions = list(
map(
int,
input("Select activation functions: ").replace(
" ", "").split(",")))
learning_rate = float(input("learning rate: "))
bulk_size = int(input("Bulk size: "))
self.move_classifier = MoveClassifier(
nr_of_training_cases=GameController.NR_OF_TRAINING_CASES,
nr_of_test_cases=GameController.NR_OF_TEST_CASES,
nr_of_nodes_in_layers=nodes_in_each_layer,
act_functions=activation_functions,
lr=learning_rate,
number_of_input_nodes=number_of_input_nodes,
number_of_output_nodes=GameController.NR_OF_OUTPUT_NODES,
bulk_size=bulk_size)
self.move_classifier.preprocessing(boards=self.move_classifier.boards,
labels=self.move_classifier.labels)
self.move_classifier.preprocessing(
boards=self.move_classifier.test_boards,
labels=self.move_classifier.test_labels)
#self.move_classifier.test_preprocessing(boards=self.move_classifier.boards, labels=self.move_classifier.labels)
if self.use_merge_input_nodes:
self.move_classifier.boards = self.move_classifier.preprocessing_row_column(
boards=self.move_classifier.boards)
self.move_classifier.test_boards = self.move_classifier.preprocessing_row_column(
boards=self.move_classifier.test_boards)
#self.move_classifier.add_extra_nodes(self.move_classifier.boards, extra_nodes)
#self.move_classifier.add_extra_nodes(self.move_classifier.test_boards, extra_test_nodes)
self.errors = []
def user_control(self):
while True:
self.action = input("Enter a command or a number to train: ")
# Test classification percentage using both the test set and training set
if self.action[0] == "t":
self.test_percentage_training_and_test_set()
# Play forever
elif self.action[0] == "s":
self.results_length = float('inf')
return
# Play 50 games, using the neural net p or a random player r
elif self.action[0] == "p" or self.action[0] == "r":
self.results_length = 50
return
# Run the grading function
elif self.action[0] == "c":
if len(self.results_from_nn_playing) < 50:
self.results_length = 50
return
else:
self.print_comparison()
elif self.action[0] == "l":
self.collect_cases = not self.collect_cases
if self.collect_cases:
self.neural_network_cases = load_cases()
self.expectimax = Expectimax()
else:
self.errors = self.move_classifier.do_training(
epochs=int(self.action), errors=self.errors)
self.test_percentage_training_and_test_set()
print("Total time elapsed: " +
str(round((time() - self.start_time) / 60, 1)) + " min")
def run_algorithm(self):
self.continuing = True
if self.game_board.is_game_over():
self.conclude_game()
return self.start_game()
current_node = State(self.game_board, self.depth)
self.move_count += 1
flat_board = current_node.board.board[3] + current_node.board.board[
2] + current_node.board.board[1] + current_node.board.board[0]
if self.collect_cases:
self.gather_case_and_result_using_expectimax(
current_node, flat_board)
if self.action[0] == "r":
chosen_move = self.choose_legal_random_move()
else:
flat_board = [flat_board]
self.move_classifier.preprocessing(boards=flat_board, labels=None)
if self.use_merge_input_nodes:
flat_board = self.move_classifier.preprocessing_row_column(
boards=flat_board)
# flat_board = self.move_classifier.add_extra_nodes([flat_board], extra_nodes)[0]
output_activations = self.move_classifier.predictor(flat_board)
chosen_move = self.choose_legal_move_from_nn(output_activations)
self.do_move(chosen_move)
self.game_board.generate_new_node()
def do_move(self, chosen_move):
if chosen_move == 0:
self.game_board.move_left()
elif chosen_move == 1:
self.game_board.move_right()
elif chosen_move == 2:
self.game_board.move_up()
elif chosen_move == 3:
self.game_board.move_down()
def conclude_game(self):
self.continuing = False
largest_tile = self.game_board.get_largest_tile()
print("Largest tile", largest_tile)
self.results.append(largest_tile)
print("Average tile", sum(self.results) / float(len(self.results)))
if self.collect_cases:
print("size of training data", len(self.neural_network_cases))
dump_cases(self.neural_network_cases)
def choose_legal_random_move(self):
while True:
r = randint(0, 3)
if self.game_board.is_move_legal(r):
return r
def choose_legal_move_from_nn(self, result):
chosen_move = None
while chosen_move == None or not self.game_board.is_move_legal(
chosen_move):
if chosen_move != None:
result[0][chosen_move] = -1
chosen_move = np.argmax(result[0])
return chosen_move
def gather_case_and_result_using_expectimax(self, current_node,
flat_board):
self.expectimax.run_expectimax(current_node, self.depth, -float("inf"),
float("inf"), None)
self.neural_network_cases[str(flat_board)] = self.expectimax.result
def welch(self, list1, list2):
params = {"results": str(list1) + " " + str(list2), "raw": "1"}
resp = requests.post('http://folk.ntnu.no/valerijf/6/', data=params)
return resp.text
def test_percentage_training_and_test_set(self):
output_activations = self.move_classifier.do_testing(
boards=self.move_classifier.test_boards)
print(
"Statistics (test set): \t\t",
self.move_classifier.check_result(
output_activations, labels=self.move_classifier.test_labels),
"%")
output_activations = self.move_classifier.do_testing(
boards=self.move_classifier.boards)
print(
"Statistics (training set):\t ",
self.move_classifier.check_result(
output_activations, labels=self.move_classifier.labels), "%")
def print_comparison(self):
print("NN results:\t", self.results_from_nn_playing)
print("Random results:\t", self.results_from_random_playing)
print("largest tiles", max(self.results_from_nn_playing),
max(self.results_from_random_playing))
print(
"average tiles",
sum(self.results_from_nn_playing) /
float(len(self.results_from_nn_playing)),
sum(self.results_from_random_playing) /
float(len(self.results_from_random_playing)))
points = self.welch(self.results_from_random_playing,
self.results_from_nn_playing)
print("points", points)
def print_commands(self):
print("Commands")
print(
"t: Test the network classification using both the test set and the training set"
)
print("l toggle case collection. Currently: ", self.collect_cases)
print("s: Run infinite times using NN")
print("p: Run 50 games using NN")
print("r: Run 50 games using a random player")
print("c: Compare the two runs of 50")
class Gui(tk.Tk):
def __init__(self, delay, diagonal=False, *args, **kwargs):
tk.Tk.__init__(self, *args, **kwargs)
self.title("2048-solver")
self.cell_width = self.cell_height = 100
self.dim = (4, 4)
self.delay=delay
screen_width = self.dim[0]*self.cell_width+1
screen_height = self.dim[1]*self.cell_height+1
self.canvas = tk.Canvas(self, width=screen_width, height=screen_height, borderwidth=0, highlightthickness=0)
self.canvas.pack(side="top", fill="both", expand="true")
self.old_nr_of_cases = 0
#self.bind_keys()
self.color_dict = self.fill_color_dict()
self.results = []
self.start_game()
#self.game_board = Game2048(board=[[0,0,0,0],[0,0,0,0],[0,0,0,0],[0,0,0,0]])
#self.board = self.game_board.board
#self.game_board.generate_new_node()
#self.draw_board()
def start_game(self):
print "avg", sum(self.results)/float(len(self.results) + 0.001)
if len(self.results) < 30:
self.game_board = Game2048(board=[[0,0,0,0],[0,0,0,0],[0,0,0,0],[0,0,0,0]])
self.board = self.game_board.board
self.game_board.generate_new_node()
self.depth = 4
self.move_count = 0
self.expectimax = Expectimax()
self.draw_board()
self.time = time()
self.run_algorithm()
else:
print self.results
print "avg", sum(self.results)/float(len(self.results))
def run_algorithm(self):
if self.game_board.open_cells_count() < 4:
self.depth = 3
else:
self.depth = 3
continuing = True
if self.game_board.is_game_over():
largest_tile = self.game_board.get_largest_tile()
print "largest tile", largest_tile
print "time elapsed: " + str(round((time() - self.time)/60, 1)) + " min"
self.results.append(largest_tile)
continuing = False
print "move count", self.move_count
return self.start_game()
current_node = State(self.game_board, self.depth)
self.move_count += 1
chosen_move = self.expectimax.run_expectimax(current_node, self.depth, -float("inf"), float("inf"), None)
expectimax_result = self.expectimax.result
flat_board = current_node.board.board[3] + current_node.board.board[2] + current_node.board.board[1] + current_node.board.board[0]
#TODO what is this? Continuing
if chosen_move == None:
Continuing = False
elif chosen_move == 0:
Continuing = False
elif chosen_move == "left":
self.game_board.move_left()
elif chosen_move == "right":
self.game_board.move_right()
elif chosen_move == "up":
self.game_board.move_up()
elif chosen_move == "down":
self.game_board.move_down()
else:
print "finished because of error in minimax chosen move"
self.game_board.generate_new_node()
self.draw_board()
if continuing:
self.after(self.delay, lambda: self.run_algorithm())
def bind_keys(self):
self.bind('<Up>', lambda event: self.move(self, self.game_board.move_up(), 0))
self.bind('<Right>', lambda event: self.move(self, self.game_board.move_right(), 1))
self.bind('<Down>', lambda event: self.move(self, self.game_board.move_down(), 2))
self.bind('<Left>', lambda event: self.move(self, self.game_board.move_left(), 3))
def move(self, event, is_moved, direction):
if is_moved:
self.game_board.generate_new_node()
self.draw_board()
self.f = open('/Users/hakon0601/Dropbox/Python/AIProg/AIProg_Module_5/2048trainingdata.txt', 'a')
board = ""
for i in range(3,-1,-1):
board += (str(self.game_board.board[i][0])) + " "
board += (str(self.game_board.board[i][1])) + " "
board += (str(self.game_board.board[i][2])) + " "
board += (str(self.game_board.board[i][3])) + " "
board += " "
board += str(direction)
self.f.write(board)
self.f.write("\n")
self.f.close()
def draw_board(self):
self.canvas.delete("all")
for y in range(self.dim[1]):
for x in range(self.dim[0]):
x1 = x * self.cell_width
y1 = self.dim[1]*self.cell_height - y * self.cell_height
x2 = x1 + self.cell_width
y2 = y1 - self.cell_height
cell_type = self.board[y][x]
text = str(self.board[y][x])
color = self.color_dict[str(self.board[y][x])]
self.canvas.create_rectangle(x1, y1, x2, y2, fill=color, tags="rect")
if cell_type != 0:
self.canvas.create_text(x1+self.cell_width/2, y1-self.cell_height/2, text=text)
def fill_color_dict(self):
color_dict = {
'0': "white",
'2': "lemon chiffon",
'4': "peach puff",
'8': "sandy brown",
'16': "dark orange",
'32': "salmon",
'64': "tomato",
'128': "khaki",
'256': "khaki",
'512': "red",
'1024': "light goldenrod",
'2048': "firebrick",
'4096': "dim grey",
'8192': "light goldenrod",
}
return color_dict
def run_game(game_class=Game2048, title='2048: In Python!', data_dir=None, **kwargs):
pygame.init()
pygame.display.set_caption(title)
AI_type = kwargs["AI_type"]
# Try to set the game icon.
try:
pygame.display.set_icon(game_class.icon(32))
except pygame.error:
# On windows, this can fail, so use GDI to draw then.
print('Consider getting a newer card or drivers.')
os.environ['SDL_VIDEODRIVER'] = 'windib'
if data_dir is None:
data_dir = user_data_dir(appauthor='Quantum', appname='2048', roaming=True)
try:
os.makedirs(data_dir)
except OSError as e:
if e.errno != errno.EEXIST:
raise
score_file_prefix = os.path.join(data_dir, '2048')
state_file_prefix = os.path.join(data_dir, '2048')
if AI_type:
score_file_prefix += '_' + AI_type
state_file_prefix += '_' + AI_type
if AI_type in ["heuristic", "rollout", "MCTS"]:
if AI_type == "rollout" and kwargs["type"] is None or kwargs["type"] == 'None':
type_str = "random"
else:
type_str = kwargs["type"]
score_file_prefix += '_' + type_str
state_file_prefix += '_' + type_str
screen = pygame.display.set_mode((game_class.WIDTH, game_class.HEIGHT))
manager = GameManager(Game2048, screen,
score_file_prefix + '.score',
state_file_prefix + '.%d.state', **kwargs)
if not AI_type:
try:
while True:
event = pygame.event.wait()
manager.dispatch(event)
for event in pygame.event.get():
manager.dispatch(event)
manager.draw()
finally:
pygame.quit()
manager.close()
else:
try:
pygame.event.set_blocked([pygame.KEYDOWN, pygame.MOUSEBUTTONUP])
manager.new_game(**kwargs)
game_scores = []
best_tiles = []
condition = True
tree = None
if AI_type in ["rollout", "MCTS"]:
num_rollouts = kwargs["num_rollouts"]
max_depth = kwargs["max_depth"]
epsilon = kwargs["epsilon"]
if epsilon < 0 or epsilon > 1:
raise ValueError("Epsilon must be in the interval [0, 1].")
if AI_type == "MCTS":
UCT = kwargs["UCT"]
tree = AI.GameTree(np.array(manager.game.grid), max_search_depth=max_depth,
num_rollouts=num_rollouts, epsilon=epsilon, UCT=UCT,
use_expert_score=kwargs["use_expert"])
while condition:
if manager.game.lost:
event = pygame.event.Event(pygame.MOUSEBUTTONUP, {"pos": manager.game.lost_try_again_pos})
game_scores.append(manager.game.score)
best_tiles.append(np.max(manager.game.grid))
print(len(game_scores))
if AI_type in ["random", "heuristic", "MCTS", "rollout", "expectimax"]:
condition = kwargs["num_games"] > len(game_scores)
elif manager.game.won == 1:
event = pygame.event.Event(pygame.MOUSEBUTTONUP, {"pos": manager.game.keep_going_pos})
elif AI_type == "random":
event = AI.random_move_event(np.array(manager.game.grid))
elif AI_type == "heuristic":
event = AI.heuristic_move_event(np.array(manager.game.grid), kwargs["type"])
elif AI_type == "rollout":
event = AI.rollouts(np.array(manager.game.grid), manager.game.score, kwargs["type"],
max_search_depth=max_depth, num_rollouts=num_rollouts, epsilon=epsilon,
use_expert_score=kwargs["use_expert"])
elif AI_type == "MCTS":
event = tree.MCTS(np.array(manager.game.grid), manager.game.score)
elif AI_type == "expectimax":
event = Expectimax(kwargs['max_depth']).get_best_move(np.array(manager.game.grid))
else:
raise ValueError("AI mode selected but invalid AI type was supplied!")
manager.dispatch(event)
manager.draw()
print("Number of games played:", len(game_scores))
print("Game Scores:")
print(game_scores)
print("Best Tiles:")
print(best_tiles)
print("Max Score:", max(game_scores))
print("Max Tile:", max(best_tiles))
print("Average Score:", stats.mean(game_scores))
finally:
if "simulate" not in kwargs:
pygame.quit()
manager.close()
if "simulate" in kwargs:
results = {
"game_scores": game_scores,
"best_tiles": best_tiles,
"max_score": max(game_scores),
"max_tile": max(best_tiles),
"avg_score": stats.mean(game_scores)
}
return results
