def run():
mdp = mdps.MazeMDP(room_size=5, num_rooms=5)
print 'online RL algorithm: '
total_rewards, V = simulate_online_RL_algorithm(mdp)
mdp.print_v(V)
learning_utils.plot_rewards(total_rewards)
print 'DP algorithm: '
simulate_MDP_algorithm(mdp)
def run():
mdp = mdps.MazeMDP(room_size=5, num_rooms=5)
print 'online RL algorithm: '
total_rewards, V = simulate_online_RL_algorithm(mdp)
mdp.print_v(V)
learning_utils.plot_rewards(total_rewards)
print 'DP algorithm: '
simulate_MDP_algorithm(mdp)
def run_keras_nnet():
mdp = mdps.MazeMDP(room_size=5, num_rooms=2)
num_episodes = 200
total_rewards, total_steps, trajectory, V = simulate_keras_online_RL_algorithm(mdp=mdp, num_episodes=num_episodes, max_iterations=100)
print 'average_reward: {}'.format(np.mean(total_rewards[num_episodes / 2:]))
print 'average_steps: {}'.format(np.mean(total_steps[num_episodes / 2]))
learning_utils.plot_rewards(total_rewards)
learning_utils.plot_rewards(total_steps)
print trajectory
mdp.print_trajectory(trajectory)
print V
mdp.print_v(V)
def run_keras_nnet():
mdp = mdps.MazeMDP(room_size=5, num_rooms=2)
num_episodes = 200
total_rewards, total_steps, trajectory, V = simulate_keras_online_RL_algorithm(
mdp=mdp, num_episodes=num_episodes, max_iterations=100)
print 'average_reward: {}'.format(np.mean(total_rewards[num_episodes /
2:]))
print 'average_steps: {}'.format(np.mean(total_steps[num_episodes / 2]))
learning_utils.plot_rewards(total_rewards)
learning_utils.plot_rewards(total_steps)
print trajectory
mdp.print_trajectory(trajectory)
print V
mdp.print_v(V)
def run_nnet():
mdp = mdps.MazeMDP(room_size=5, num_rooms=2)
total_rewards, total_losses = simulate_symbolic_online_RL_algorithm(mdp=mdp, num_episodes=700, max_iterations=100)
learning_utils.plot_rewards(total_rewards)
learning_utils.plot_rewards(total_losses)
def run_nnet():
mdp = mdps.MazeMDP(room_size=5, num_rooms=2)
total_rewards, total_losses = simulate_symbolic_online_RL_algorithm(
mdp=mdp, num_episodes=700, max_iterations=100)
learning_utils.plot_rewards(total_rewards)
learning_utils.plot_rewards(total_losses)
