QLearnerAgentClass.py 文件源码

def epsilon_greedy_q_policy(self, state):
        '''
        Args:
            state (State)

        Returns:
            (str): action.
        '''
        # Policy: Epsilon of the time explore, otherwise, greedyQ.
        if numpy.random.random() > self.epsilon:
            # Exploit.
            action = self.get_max_q_action(state)
        else:
            # Explore
            action = numpy.random.choice(self.actions)

        return action