def revealed_q():
width = 3
height = 3
num_mines = 1
sim = Minesweeper(num_mines, width, height)
#features = sim.reset()
Q = np.ones((sim.num_states, sim.num_actions))
policy, Q, steps, reward, r, num_revealed = q_learning(
sim, gamma=gamma,alpha = alpha, epsilon=epsilon,
num_episodes=1, initial_Q=Q)
board_covered = num_revealed / (width * height)
return board_covered
def revealed_mc():
width = 3
height = 3
num_mines = 1
sim = Minesweeper(num_mines, width, height)
#features = sim.reset()
Q = np.ones((sim.num_states, sim.num_actions))
policy, Q, steps,reward, num_rev = monte_carlo_iterative_optimisation(
sim, gamma, epsilon, alpha, num_episodes = 1, max_steps=20,
initial_Q=Q, default_value=0)
#print(r, reward)
board_covered = num_rev / (width * height)
return board_covered
def run_q():
width = 3
height = 3
num_mines = 1
sim = Minesweeper(num_mines, width, height)
# intial Q table
Q = np.ones((sim.num_states, sim.num_actions))
# running simulatino with q learning
policy, Q, steps, reward, r, num_revealed = q_learning(sim,
gamma=gamma,
alpha=alpha,
epsilon=epsilon,
num_episodes=1,
initial_Q=Q)
return reward
def steps_sarsa():
width = 3
height = 3
num_mines = 1
sim = Minesweeper(num_mines, width, height)
# initial Q table
Q = np.ones((sim.num_states, sim.num_actions))
# running simulatino with sarsa
policy, Q, steps, reward, r, num = sarsa(sim,
gamma=gamma,
alpha=alpha,
epsilon=epsilon,
num_episodes=1,
initial_Q=Q)
return steps
def steps_mc():
width = 3
height = 3
num_mines = 1
sim = Minesweeper(num_mines, width, height)
#features = sim.reset()
Q = np.ones((sim.num_states, sim.num_actions))
policy, Q, steps, reward, num_rev = monte_carlo_iterative_optimisation(
sim,
gamma,
epsilon,
alpha,
num_episodes=1,
max_steps=40,
initial_Q=Q,
default_value=0)
return steps
def main():
width = 3
height = 3
num_mines = 1
sim = Minesweeper(num_mines, width, height)
features = sim.reset()
action = np.random.choice([0, 1])
qweights = np.zeros((2, 80))
for i in range(80):
qweights[1] = 1
#print("\tAction choice is: %d (meaning %s)" \
# % (action, sim.action_names[action],) )
#rewards_arr = []
sim = Minesweeper(num_mines, width, height)
features = sim.reset()
action = np.random.choice([0, 1])
while sim.is_terminal() == False:
features, reward, num_revealed = sim.next(action)
return reward
def random():
width = 3
height = 3
num_mines = 1
sim = Minesweeper(num_mines, width, height)
features = sim.reset()
action = np.random.choice([0,1])
qweights = np.zeros((2,80))
for i in range(80):
qweights[1] = 1
sim = Minesweeper(num_mines, width, height)
action = np.random.choice([0,1])
while sim.is_terminal() == False:
features, reward, num_revealed = sim.next(action)
board_covered = num_revealed / (width * height)
return board_covered