def return_results(self, temp):
''' Process the output at the root node '''
counts = np.array([child_action.n for child_action in self.root.child_actions])
Q = np.array([child_action.Q for child_action in self.root.child_actions])
pi_target = stable_normalizer(counts, temp)
V_target = np.sum((counts / np.sum(counts)) * Q)[None]
return self.root.index.flatten(), pi_target, V_target
def return_results(self, temp):
""" Process the output at the root node """
counts = np.array(
[child_action.n for child_action in self.root.child_actions])
q = np.array(
[child_action.q for child_action in self.root.child_actions])
pi_target = stable_normalizer(counts, temp)
v_target = np.sum((counts / np.sum(counts)) * q)[None]
return self.root.index, pi_target, v_target
def return_results(self, temp, on_visits=False):
""" Process the output at the root node """
counts = np.array(
[child_action.n for child_action in self.root.child_actions])
Q = np.array(
[child_action.Q for child_action in self.root.child_actions])
if on_visits:
pi_target = stable_normalizer(counts, temp)
else:
pi_target = max_Q(Q)
V_target = np.sum((counts / np.sum(counts)) * Q)[None]
return self.root_signature, pi_target, V_target