def main(args):
length = args.l
breadth = args.b
learning_rate = args.lr
discount_rate = args.dr
min_explore_rate = args.p
max_explore_rate = args.q
decay_factor = args.d
num_episodes = args.ep
num_iterations = args.iter
board = grid.draw_grid(length, breadth)
# board = [['S', '-', '*', '*', '-'], ['*', '-', '*', '*', '-'], ['-', '-', '*', '-', '-'], ['-', '*', '-', '-', '-'], ['-', '-', '-', '*', 'G']]
print("Board -:\n\n", board)
q_table = np.zeros((length * breadth, len(actions)))
explore_rate = max_explore_rate
i = 0
while i < num_episodes:
j = 0
completed = False
x = y = 0 # Start State
while j < num_iterations and completed == False:
current_state = (x, y)
action = choose_action(q_table, current_state, explore_rate,
breadth)
next_state = perform_action(current_state, action, length, breadth)
if board[next_state[0]][next_state[1]] == 'G':
completed = True
reward = get_reward(board, next_state)
x_new, y_new = next_state
q_table[x * breadth + y][actions.index(action)] = q_table[
x * breadth + y][actions.index(action)] + learning_rate * (
reward +
discount_rate * q_table[x_new * breadth + y_new].max() -
q_table[x * breadth + y][actions.index(action)])
x, y = x_new, y_new
explore_rate = min_explore_rate + (max_explore_rate -
min_explore_rate) * np.exp(
-decay_factor *
(i * num_iterations + j))
j = j + 1
i = i + 1
print("Episode {} : Iterations : {}".format(i, j))
print("\nFinal Q-Table -:\n\n", q_table)
def test_lsd_consistant(fit=True):
import matplotlib.pyplot as plt
image = tf.placeholder(tf.uint8, shape=[None, None, 3])
N = [2, 3]
path = '/media/george/1a4f4334-123f-430d-8a2b-f0c0fa401c75/data/alchohol/alchohol_104.jpeg'
lines = lsd(image, N, fit=fit)
sess = tf.InteractiveSession()
_image = cv2.cvtColor(cv2.imread(path, cv2.IMREAD_COLOR),
cv2.COLOR_BGR2RGB)
_image1 = draw_grid(_image.copy(), N)
plt.imshow(_image1)
plt.show()
_lines = sess.run(lines, feed_dict={image: _image})
for i in range(_lines.shape[0]):
line = _lines[i]
if not check_in_single_grid(line, N):
print(line)
menu.draw_character_builder(screen)
pygame.display.flip()
elif settings.game_state == constants.GAMEPLAY:
surface_messages.fill(constants.COLOR_BG)
surface_grid.fill(constants.COLOR_BG)
surface_inv.fill(constants.COLOR_BG)
surface_equip.fill(constants.COLOR_BG)
# Draw messages
text.draw_messages(surface_messages)
screen.blit(surface_messages, (10, 10))
# Draw grid
grid.draw_grid(settings.grid, surface_grid)
for v in settings.tiles:
v.draw(surface_grid)
for v in settings.item_stacks:
v.draw(surface_grid)
for v in settings.monsters:
v.draw(surface_grid)
screen.blit(surface_grid, (10, 110))
# Draw inv
pygame.draw.rect(surface_inv, (0, 0, 0),
pygame.Rect(170, 0, 5, 40))
if is_inv:
inv.draw_inv(settings.player.inv, surface_inv)
else:
placed.add(coming_pipes[2]((-2, 1)))
placed.add(coming_pipes[1]((-2, 2)))
placed.add(coming_pipes[0]((-2, 3)))
# Fyller surface med farge (RGB)
surface.fill((255, 255, 255))
# Listen har en update() funksjon. Den kaller i tur og orden update() funksjonen
# på objektene vi har lagt til i elements listen.
# eks. For sprite in elements.sprites:
# sprite.update()
# Selv om spritene er av forskjellig type, så vil en kunne kalle update() uten å sjekke type.
# Det er det som menes med "polymorfi" i objektorientertprogramering.
planned.update()
placed.update()
draw_grid(surface)
# elements.draw() går gjennom alle objektene i listen.
# Fuksjonen bruker så image og rect egenskapene fra spritene for å tegne dem på "surface"
planned.draw(surface)
placed.draw(surface)
# Her blir surface tegnet på skjerm
screen.blit(surface, (0, 0))
# Disse funksjonene forteller pygame at "skjermen" skal oppdateres med innholdet i "screen"
pygame.display.flip()
pygame.display.update()
# Her tar programmet en pause for at det skal kjøre med en hastighet på frames pr. second.
# FPS er en "konstant" vi har opprettet i settings modulen