directory = plot_dir,
file_name = "3D_exp_MC_" + str(decks) + "_decks.png", show = False)
"""
"""
# Extended state Q-learning
Q_conv = rl.convert_to_sum_states(Q, env)
V_conv = rl.convert_to_value_function(Q_conv)
V_conv_filt = rl.fill_missing_sum_states(rl.filter_states(V_conv))
pl.plot_value_function(V_conv_filt,
title = "Expanded state, " + str(decks) + " decks",
directory = plot_dir,
file_name = "3D_exp_" + str(decks) + "_decks.png")
"""
# Likewise make 3D plots for sumQ
V_sum = rl.convert_to_value_function(sumQ)
V_sum_filt = rl.fill_missing_sum_states(rl.filter_states(V_sum))
pl.plot_value_function(V_sum_filt,
title="Sum state, " + str(decks) + " decks",
directory=plot_dir,
file_name="3D_sum_" + str(decks) + "_decks.png")
"""
# create line plots
env_types = ["hand_MC", "hand", "sum"]
fig, lgd = pl.plot_avg_reward_episode(directory, env_types, [str(decks)])
fig.savefig("{}/avgReturnEp_ndeck{}.png".format(plot_dir, decks),
bbox_extra_artists=(lgd,), bbox_inches='tight')
matplotlib.pyplot.close(fig)
"""
keys = list(sumQ.keys())
warmup=warmup)
time_to_completion_sum = time.time() - start_time_sum
print("Number of explored states (sum states): " + str(len(sumQ)))
print("Cumulative avg. reward = " + str(sum_avg_reward))
print(
"Training time: \n " +
"Expanded state space MC: {} \n Expanded state space: {} \n Sum state space: {}"
.format(time_to_completion_MC, time_to_completion_expanded,
time_to_completion_sum))
# Convert Q (extended state) to sum state representation and make 3D plots
# Extended state MC-learning
Q_conv_MC = rl.convert_to_sum_states(Q_MC, env)
V_conv_MC = rl.convert_to_value_function(Q_conv_MC)
V_conv_filt_MC = rl.fill_missing_sum_states(
rl.filter_states(V_conv_MC))
pl.plot_value_function(
V_conv_filt_MC,
title="Expanded state MC, " + str(decks) + " decks",
directory=plot_dir,
file_name="3D_exp_MC_" + str(decks) + "_decks.png")
# Extended state Q-learning
Q_conv = rl.convert_to_sum_states(Q, env)
V_conv = rl.convert_to_value_function(Q_conv)
V_conv_filt = rl.fill_missing_sum_states(rl.filter_states(V_conv))
pl.plot_value_function(V_conv_filt,
title="Expanded state, " + str(decks) +
" decks",
