def choose_action(self, q_values):
self.update_parameters()
exp_q_values = np.exp(q_values / (self.tau + 1e-2))
weights = dict()
for idx, val in enumerate(exp_q_values):
weights[idx] = val
action = learning_utils.weightedRandomChoice(weights)
return action
def choose_action(self, q_values):
self.update_parameters()
exp_q_values = np.exp(q_values / (self.tau + 1e-2))
weights = dict()
for idx, val in enumerate(exp_q_values):
weights[idx] = val
action = learning_utils.weightedRandomChoice(weights)
return action
def get_action(self, state):
self.num_iters += 1
# if self.tau > 1e-9:
# self.tau *= .9999
# print self.tau
q_values = np.array([self.getQ(state, action) for action in self.actions])
exp_q_values = np.exp(q_values / (self.tau + 1e-2))
weights = dict()
for idx, val in enumerate(exp_q_values):
weights[idx] = val
action_idx = learning_utils.weightedRandomChoice(weights)
action = self.actions[action_idx]
return action