def runCrossEntropyMethod(initMeanV, initSdV):
n = 30
rho = 0.1
k = int(rho * n)
meanV = initMeanV
sdV = initSdV
for i in xrange(11, 101):
meanV, sdV = runCrossEntropyIteration(n, k, meanV, sdV)
print strftime("%Y-%m-%d %H:%M:%S", gmtime())
print i, 'mean', meanV
print i, 'sd', sdV
if i % 10 == 0:
game.writeWeights('weights_ce{}.tetris'.format(i), meanV)
game.writeWeights('sd_ce{}.tetris'.format(i), sdV)
if successor.isWin():
successor = None
elif successor.isLose():
# reward = -100
successor = None
output.append((successor, prob, reward))
return output
def discount(self):
return self.tetris.discount()
if __name__ == "__main__":
# weights = TetrisGame.readWeights('weightsThiery.tetris')
#weights = [0.10000000000000001, 0.5, -44.892192895842715, -88.868713917352991, -9.0845044328524125, 99.48473353935913, -0.10000000000000001, -0.10000000000000001, -46.3552758541178784, -44.894981251780582, -49.903753991739844, -48.776916252514777, -56.009249265806467, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
# weights = [random.random() for _ in xrange(22)]
weights = TetrisGame.readWeights('weights_ql400.tetris')
# evaluator = AdvancedEvaluator(weights)
evaluator = AdhocEvaluator(weights)
baseSeq = [0,1,2,3,4,5,6]*100
# random.seed(21)
random.shuffle(baseSeq);
# state: (board, new piece)
# action: (newGrid, reward)
def featureExtractor(state, action):
newState = state.generateSuccessor(0, action)
output = evaluator.featureExtractor(newState)
return output
model = TetrisGameMDP(TetrisMDP(), ExpectimaxTetrisAgent(0, 1, evaluator), FinitePieceGenerator(baseSeq))
print 'std', std
print '5%', percent5
print '95%', percent95
return mean, std, percent5, percent95
def printArr(arr):
print '=========='
for a in arr:
print a
if __name__ == '__main__':
# weightFile = 'weights_ql500.tetris'
results_lines = []
results_scores = []
for i in xrange(9,21):
weightFile = 'weights_ce{}.tetris'.format(i)
weights = game.readWeights(weightFile)
linesV, scoresV = simulateGame(weights, 10)
print '===lines==='
results_lines.append(analyze(linesV))
print '===scores==='
results_scores.append(analyze(scoresV))
results1 = zip(*results_lines)
results2 = zip(*results_scores)
print len(results1), len(results2)
for result in results1:
printArr(result)
for result in results2:
printArr(result)
pygame.display.set_caption("Tetris")
clock = pygame.time.Clock()
# Create game_board surface for the tetris grid
game_board_size = [i * tetris_pixel_size for i in (10, 24)]
surface_board = pygame.Surface(game_board_size)
surface_saved = pygame.Surface((4 * tetris_pixel_size, 4 * tetris_pixel_size))
surface_upcoming = pygame.Surface(((2 * tetris_pixel_size), ((4 * tetris_pixel_size * 5) + upcoming_tetris_display_gap * 5)))
surface_upcoming.fill((0, 100, 100))
# Create font for text
text_font = pygame.font.SysFont('Comic Sans MS', 16)
# Create tetris game object
tetrisGame = TetrisGame.TetrisGame()
run = True
while run:
# Get and process events, including keypress
for event in pygame.event.get():
if event.type == pygame.QUIT:
run = False
if event.type == pygame.KEYDOWN:
# this method of getting keypresses does not repeat
if event.key == pygame.K_DOWN:
tetrisGame.move_drop_soft(1)
if event.key == pygame.K_UP:
tetrisGame.move_rotate(1)
if event.key == pygame.K_RIGHT:
tetrisGame.move_horizontal(1)
