def test_training_TD_for_gridworld(model_class, train=True):
gridworld = GridWorld()
if train:
policy, model = rl.train_reinforcement_strategy_temporal_difference(
epochs=50000, game_obs=gridworld, model_class=model_class
)
rl.test_policy(gridworld)
def test_training_TD_for_blackjack(model_class):
blackjack = BlackJack()
policy, model = rl.train_reinforcement_strategy_temporal_difference(
epochs=5000, game_obs=blackjack, model_class=model_class)
df = pd.DataFrame(policy).T
df.columns = ['player_value', 'dealer_value', 'decision', 'score']
policy_Q_table = df.pivot('player_value', 'dealer_value')['decision']
print policy_Q_table
policy_Q_score = df.pivot('player_value', 'dealer_value')['score']
print policy_Q_score
# Add ipython notebook 3D ghaph
# Test policy
rl.test_policy(blackjack)
return policy, model
def test_training_TD_for_blackjack(model_class):
blackjack = BlackJack()
policy, model = rl.train_reinforcement_strategy_temporal_difference(
epochs=5000, game_obs=blackjack, model_class=model_class
)
df = pd.DataFrame(policy).T
df.columns = ["player_value", "dealer_value", "decision", "score"]
policy_Q_table = df.pivot("player_value", "dealer_value")["decision"]
print policy_Q_table
policy_Q_score = df.pivot("player_value", "dealer_value")["score"]
print policy_Q_score
# Add ipython notebook 3D ghaph
# Test policy
rl.test_policy(blackjack)
return policy, model
def test_training_TD_for_gridworld(model_class, train=True):
gridworld = GridWorld()
if train:
policy, model = rl.train_reinforcement_strategy_temporal_difference(
epochs=50000, game_obs=gridworld, model_class=model_class)
rl.test_policy(gridworld)
