hiv.py 文件源码

def calc_reward(self, action=0, state=None, **kw ):
        """Calculate the reward for the specified transition."""
        eps1, eps2 = self.eps_values_for_actions[action]
        if state is None:
            state = self.observe()
        if self.logspace:
            T1, T2, T1s, T2s, V, E = 10**state
        else:
            T1, T2, T1s, T2s, V, E = state
        # the reward function penalizes treatment because of side-effects
        reward = -0.1*V - 2e4*eps1**2 - 2e3*eps2**2 + 1e3*E
        # Constrain reward to be within specified range
        if np.isnan(reward):
            reward = -self.reward_bound
        elif reward > self.reward_bound:
            reward = self.reward_bound
        elif reward < -self.reward_bound:
            reward = -self.reward_bound
        return reward