def alpha_beta_move(board,active_turn,depth,alpha = 2):
swap_dict = {1:-1,-1:1}
dummy_board = np.zeros((6,7))
dummy_board[:] = board[:]
options = cccc.available_moves(board)
random.shuffle(options)
if len(options) == 1:
dummy_board[np.where(dummy_board[:,options[0]]==0)[0][-1],options[0]] = active_turn
if cccc.winner(dummy_board):
return (1,options[0]+1)
else:
return (0,options[0]+1)
if depth ==0:
return (0, options[np.random.randint(len(options))]+1)
best_value = -2
candidate_move = None
for x in options:
height = np.where(dummy_board[:,x]==0)[0][-1]
dummy_board[height,x] = active_turn
if cccc.winner(dummy_board):
return (1, x+1)
(opp_value,opp_move) = alpha_beta_move(dummy_board,swap_dict[active_turn],depth-1,-best_value)
if -opp_value > best_value:
candidate_move = x+1
best_value = -opp_value
if -opp_value >= alpha:
#print (options, x, best_value, alpha)
break
dummy_board[height,x] = 0
return (best_value, candidate_move)
def alpha_beta_move(board,
active_turn,
depth,
evaluation=lambda x: 0,
alpha=2):
swap_dict = {1: -1, -1: 1}
dummy_board = np.copy(board)
dummy_board = dummy_board.reshape((6, 7))
options = cccc.available_moves(dummy_board)
random.shuffle(options)
if len(options) == 1:
dummy_board[np.where(dummy_board[:, options[0]] == 0)[0][-1],
options[0]] = active_turn
if cccc.winner(dummy_board):
return (1, options[0] + 1)
else:
return (0, options[0] + 1)
if depth == 0:
best_value = -2
for x in options:
height = np.where(dummy_board[:, x] == 0)[0][-1]
dummy_board[height, x] = active_turn
eval_board = evaluation(dummy_board * active_turn)
if eval_board > best_value:
best_value = eval_board
candidate_move = x + 1
dummy_board[height, x] = 0
return (best_value, candidate_move)
best_value = -2
candidate_move = None
for x in options:
height = np.where(dummy_board[:, x] == 0)[0][-1]
dummy_board[height, x] = active_turn
if cccc.winner(dummy_board):
return (1, x + 1)
(opp_value,
opp_move) = alpha_beta_move(dummy_board, swap_dict[active_turn],
depth - 1, evaluation, -best_value)
if -opp_value > best_value:
candidate_move = x + 1
best_value = -opp_value
if -opp_value >= alpha:
#print (options, x, best_value, alpha)
break
dummy_board[height, x] = 0
return (best_value, candidate_move)
def play_cccc():
print "requesting move"
player_index_dict = {-1: 1, 1: 0}
player = int(request.form.get("player"))
depths = map(int, request.form.get("depths").split(','))
print "the depths are ", depths
board = request.form.get("board")
board = board.split(",")
board = [int(x) for x in board]
board = np.array(board).reshape((6, 7))
print "the board is "
print board
types = request.form.get("types")
types = map(lambda x: x.replace("\"", ""), types.split(","))
evals = request.form.get("evals").split(",")
print "the eval method is ", evals[player_index_dict[player]]
if fc4.game_over(np.copy(board).reshape((6, 7))):
finished = cccc.winner(board.reshape((6, 7)))
return jsonify(finished=finished)
else:
finished = -2
if evals[player_index_dict[player]] == 'nn':
evaluation = fc4.net_value
else:
evaluation = lambda x: 0
#evaluation
print "the control methods are ", types
# print types, player, fc4.game_over(np.copy(board)), evals, depths[player_index_dict[player]]
if types[player_index_dict[player]] == 'remote' and not fc4.game_over(
np.copy(board)):
move = fc4.alpha_beta_move(board.reshape((6, 7)),
player,
depth=depths[player_index_dict[player]],
evaluation=evaluation)[1]
print "the next move is ", move
fc4.update_move(board, move, player)
print "the board is "
print board
player *= -1
board = board.reshape(42)
print 'next move is', move
if fc4.game_over(np.copy(board).reshape((6, 7))):
finished = cccc.winner(board.reshape((6, 7)))
else:
finished = -2
return jsonify(move=move, player=-1 * player, finished=finished)
def play_cccc():
print "requesting move"
player_index_dict = {-1:1,1:0}
player = int(request.form.get("player"))
depths = map(int,request.form.get("depths").split(','))
print "the depths are ", depths
board = request.form.get("board")
board = board.split(",")
board = [int(x) for x in board]
board = np.array(board).reshape((6,7))
print "the board is "
print board
types = request.form.get("types")
types = map(lambda x: x.replace("\"",""),types.split(","))
evals = request.form.get("evals").split(",")
print "the eval method is ",evals[player_index_dict[player]]
if fc4.game_over(np.copy(board).reshape((6,7))):
finished = cccc.winner(board.reshape((6,7)))
return jsonify(finished = finished)
else:
finished = -2
if evals[player_index_dict[player]] == 'nn':
evaluation = fc4.net_value
else:
evaluation = lambda x: 0
#evaluation
print "the control methods are ", types
# print types, player, fc4.game_over(np.copy(board)), evals, depths[player_index_dict[player]]
if types[player_index_dict[player]] == 'remote' and not fc4.game_over(np.copy(board)):
move = fc4.alpha_beta_move(board.reshape((6,7)),
player,
depth = depths[player_index_dict[player]],
evaluation = evaluation)[1]
print "the next move is ",move
fc4.update_move(board,move,player)
print "the board is "
print board
player *= -1
board = board.reshape(42)
print 'next move is', move
if fc4.game_over(np.copy(board).reshape((6,7))):
finished = cccc.winner(board.reshape((6,7)))
else:
finished = -2
return jsonify(move=move, player = -1*player, finished = finished)
def get_max_future(future_board,value_fun):
options = cccc.available_moves(future_board)
dummy_board = np.copy(future_board)
move_values = np.zeros(7)
for move in options:
dummy_board = np.copy(future_board)
dummy_board[np.where(dummy_board[:,move]==0)[0][-1],move] = -1
# dummy_board = dummy_board.reshape(1,42)
if cccc.winner(dummy_board):
move_values[move] = cccc.winner(dummy_board)
else:
reshapable = np.copy(dummy_board)
reshapable = reshapable.reshape(1,42)
move_values[move] = value_fun(reshapable)
available_move_values = np.array([move_values[move] for move in options])
dummy_board = np.copy(future_board)
options_index = np.argmin(available_move_values)
dummy_board[np.where(dummy_board[:,options[options_index]]==0)[0][-1],options[options_index]] = -1
return np.amin(available_move_values), dummy_board
def alpha_beta_move(board,active_turn,depth,evaluation = lambda x: 0,alpha = 2):
swap_dict = {1:-1,-1:1}
dummy_board = np.copy(board)
dummy_board = dummy_board.reshape((6,7))
options = cccc.available_moves(dummy_board)
random.shuffle(options)
if len(options) == 1:
dummy_board[np.where(dummy_board[:,options[0]]==0)[0][-1],options[0]] = active_turn
if cccc.winner(dummy_board):
return (1,options[0]+1)
else:
return (0,options[0]+1)
if depth ==0:
best_value = -2
for x in options:
height = np.where(dummy_board[:,x]==0)[0][-1]
dummy_board[height,x] = active_turn
eval_board = evaluation(dummy_board*active_turn)
if eval_board > best_value:
best_value = eval_board
candidate_move = x + 1
dummy_board[height,x] = 0
return (best_value, candidate_move)
best_value = -2
candidate_move = None
for x in options:
height = np.where(dummy_board[:,x]==0)[0][-1]
dummy_board[height,x] = active_turn
if cccc.winner(dummy_board):
return (1, x+1)
(opp_value,opp_move) = alpha_beta_move(dummy_board,swap_dict[active_turn],depth-1,evaluation,-best_value)
if -opp_value > best_value:
candidate_move = x+1
best_value = -opp_value
if -opp_value >= alpha:
#print (options, x, best_value, alpha)
break
dummy_board[height,x] = 0
return (best_value, candidate_move)
def go():
if request.method == 'GET':
board = app.board
player = app.player
print "HOWDY"
return render_template('connect_four.html', board = board, cplayer = player, finished = -2 )
if request.method == 'POST':
player = int(request.form.get("player"))
board = request.form.get("board")
board = board.split(",")
#print board
board = [int(x) for x in board]
board = np.array(board)
#print board,player
if game_over(np.copy(board)):
if cccc.winner(board.reshape((6,7)))==1:
print '1'
return render_template('connect_four.html', board = list(board), cplayer = player, finished = 1)
if cccc.winner(board.reshape((6,7))) ==0:
print '0'
return render_template('connect_four.html', board = list(board), cplayer = player, finished = 0)
if cccc.winner(board.reshape((6,7))) == -1:
print '-1'
return render_template('connect_four.html', board = list(board), cplayer = player, finished = -1)
while not game_over(board):
if player == -1:
_,move = alpha_beta_move(board,player,depth=2, evaluation = net_value)
#print move
#print game_over(board)
board = board.reshape((6,7))
board[np.where(board[:,move-1]==0)[0][-1],move-1] = active_turn = player
player = -1*player
board = board.reshape(42)
#print board,player
elif player == 1:
#print board, player
return render_template('connect_four.html', board = list(board), cplayer = player,finished=-2)
if game_over(np.copy(board)):
if cccc.winner(board.reshape((6,7)))==1:
print '1'
return render_template('connect_four.html', board = list(board), cplayer = player, finished = 1)
if cccc.winner(board.reshape((6,7))) ==0:
print '0'
return render_template('connect_four.html', board = list(board), cplayer = player, finished = 0)
if cccc.winner(board.reshape((6,7))) == -1:
print '-1'
return render_template('connect_four.html', board = list(board), cplayer = player, finished = -1)
def play_cccc():
print "requesting move"
player_index_dict = {-1:1,1:0}
player = int(request.form.get("player"))
depths = map(int,request.form.get("depths").split(','))
print "the depths are ", depths
board = request.form.get("board")
board = board.split(",")
board = [int(x) for x in board]
board = np.array(board).reshape((6,7))
move =0
print "the board is "
print board
types = request.form.get("types")
types = map(lambda x: x.replace("\"",""),types.split(","))
evals = request.form.get("evals").split(",")
print "the eval method is ",evals[player_index_dict[player]]
try:
print request.form.get("overcheck")
overcheck = request.form.get("overcheck")=="true"
print overcheck
if overcheck:
if cccc.game_over(board):
finished = cccc.winner(board.reshape((6,7)))
else:
finished = -2
winners = map(list,zip(*fc4.winning_squares(board)))
if not winners:
winners = [[],[]]
print winners
return jsonify(finished = finished, y=winners[0],x=winners[1])
except:
print "failed"
if fc4.game_over(np.copy(board).reshape((6,7))):
finished = cccc.winner(board.reshape((6,7)))
winners = map(list,zip(*fc4.winning_squares(board)))
if not winners:
winners = [[],[]]
print winners
return jsonify(finished = finished, y=winners[0],x=winners[1])
else:
winners = []
finished = -2
if evals[player_index_dict[player]] == 'nn':
evaluation = fc4.sym_net_value
else:
evaluation = lambda x: 0
#evaluation
print "the control methods are ", types
# print types, player, fc4.game_over(np.copy(board)), evals, depths[player_index_dict[player]]
if types[player_index_dict[player]] == 'remote' and not fc4.game_over(np.copy(board)):
move = fc4.alpha_beta_move(board.reshape((6,7)),
player,
depth = depths[player_index_dict[player]],
evaluation = evaluation)[1]
print "the next move is ",move
fc4.update_move(board,move,player)
print "the board is "
print board
player *= -1
board = board.reshape(42)
# print 'next move is', move
if fc4.game_over(np.copy(board).reshape((6,7))):
finished = cccc.winner(board.reshape((6,7)))
winners = map(list,zip(*fc4.winning_squares(board)))
if not winners:
winners = [[],[]]
print winners
return jsonify(move=move, player = -1*player, finished = finished, y=winners[0],x=winners[1])
else:
finished = -2
winners = []
return jsonify(move=move, player = -1*player, finished = finished)
def game_over(board):
board = np.array(board).reshape((6,7))
return cccc.winner(board) or cccc.is_full(board)
def game_over(board):
board = np.array(board).reshape((6, 7))
return cccc.winner(board) or cccc.is_full(board)
def go():
if request.method == 'GET':
board = app.board
player = app.player
print "HOWDY"
return render_template('connect_four.html',
board=board,
cplayer=player,
finished=-2)
if request.method == 'POST':
player = int(request.form.get("player"))
board = request.form.get("board")
board = board.split(",")
#print board
board = [int(x) for x in board]
board = np.array(board)
#print board,player
if game_over(np.copy(board)):
if cccc.winner(board.reshape((6, 7))) == 1:
print '1'
return render_template('connect_four.html',
board=list(board),
cplayer=player,
finished=1)
if cccc.winner(board.reshape((6, 7))) == 0:
print '0'
return render_template('connect_four.html',
board=list(board),
cplayer=player,
finished=0)
if cccc.winner(board.reshape((6, 7))) == -1:
print '-1'
return render_template('connect_four.html',
board=list(board),
cplayer=player,
finished=-1)
while not game_over(board):
if player == -1:
_, move = alpha_beta_move(board,
player,
depth=2,
evaluation=net_value)
#print move
#print game_over(board)
board = board.reshape((6, 7))
board[np.where(board[:, move - 1] == 0)[0][-1],
move - 1] = active_turn = player
player = -1 * player
board = board.reshape(42)
#print board,player
elif player == 1:
#print board, player
return render_template('connect_four.html',
board=list(board),
cplayer=player,
finished=-2)
if game_over(np.copy(board)):
if cccc.winner(board.reshape((6, 7))) == 1:
print '1'
return render_template('connect_four.html',
board=list(board),
cplayer=player,
finished=1)
if cccc.winner(board.reshape((6, 7))) == 0:
print '0'
return render_template('connect_four.html',
board=list(board),
cplayer=player,
finished=0)
if cccc.winner(board.reshape((6, 7))) == -1:
print '-1'
return render_template('connect_four.html',
board=list(board),
cplayer=player,
finished=-1)
def alpha_beta_move(board, turn, depth = 0, alpha = (-inf,-inf), beta = (inf,inf), evaluation = lambda x: 0):
dummy_board = np.copy(board) # we don't want to change the board state
swap_player = {1:-1,-1:1} # So we can change whose turn
options = cccc.available_moves(board) # get legal moves
random.shuffle(options) # should inherit move order instead of randomizing
# if len(options) == 1:
# update_move(board,options[0])
# if cccc.winner(dummy_board):
# return (inf,options[0])
# else:
# return (0,options[0])
best_value = (-inf,-inf)
if not options:
print board, cccc.game_over(board)
print 'oops, no available moves'
cand_move = options[0]
if depth == 0:
for x in options:
update_move(dummy_board,x,turn)
op_value = (evaluation(dummy_board*swap_player[turn]) , depth)
if tuple(-1 * el for el in op_value) > best_value:
cand_move = x
best_value = tuple(-1 * el for el in op_value)
alpha = max(alpha, best_value)
# print depth,-op_value, best_value, cand_move,alpha,beta
if alpha >= beta:
# print 'pruned'
break #alpha-beta cutoff
unupdate_move(dummy_board,x)
else:
for x in options:
# dummy_board = np.copy(board)
# height= np.where(board[:,x]==0)[0][-1] #connect four only
# dummy_board[height, x] = turn
update_move(dummy_board,x,turn)
if cccc.winner(dummy_board): #should check over and tied too
return((inf,depth), x)
if cccc.is_full(dummy_board): #This assumes you can't lose on your turn
return((0,depth) , x)
op_value,_ = alpha_beta_move( dummy_board,
swap_player[turn],
depth-1,
alpha = tuple(-1 * el for el in beta),
beta = tuple(-1 * el for el in alpha),
evaluation = evaluation)
if tuple(-1 * el for el in op_value) > best_value:
cand_move = x
best_value = tuple(-1 * el for el in op_value)
alpha = max(alpha, best_value)
# print depth,-op_value, best_value, cand_move,alpha,beta
if alpha >= beta:
# print 'pruned'
break #alpha-beta cutoff
unupdate_move(dummy_board,x)
# dummy_board[height, x] = 0
return (best_value, cand_move)
def game_over(board):
return cccc.winner(board) or cccc.is_full(board)
game_length = len(result_O.log)
input_list = [-1*board_list[2*i+1] for i in range(game_length/2)]
output_list = [-1*board_list[2*i+2] for i in range(game_length/2)]
move_list = [result_O.log[2*i+1] for i in range(game_length/2)]
for epoch in range(train_duration):
t1 = time.clock()
if len(input_history) > minimum_data:
target_history = np.zeros(len(output_history))
print 'Creating Targets for {} data points'.format(len(output_history))
print '\n'
t3 = time.clock()
for i,item in enumerate(output_history):
output_state = np.copy(output_history[i])
if cccc.winner(output_state) or cccc.is_full(output_state):
target_history[i] = cccc.winner(output_state)
else:
#minus because the future term is in terms of the valuation for the player, and we need a target for the
#opponent
# targets[i] = (1-future_discount) * reward_state + future_discount * get_max_future(
#output_state,value_fun)
#targets = np.array(targets).reshape(BATCH_SIZE,1)
#temporal difference method
target_history[i]= 0
current_state = np.copy(output_state)
depth = 0
player = 1
