4222b5c6a93692293b471969377cb371bd907bd7,mushroom/algorithms/batch_td.py,FQI,_partial_fit,#FQI#Any#Any#,69

Before Change

        if y is None:
            target = reward
        else:
            max_q, _ = max_QA(next_state, absorbing, self.approximator)
            target = reward + self._gamma * max_q

        sa = [state, action]

After Change

            target = reward
        else:
            q = self.approximator.predict_all(next_state)
            if np.any(absorbing):
                q *= 1 - absorbing.reshape(-1, 1)

            max_q = np.max(q, axis=1)
            target = reward + self._gamma * max_q

        sa = [state, action]
        self.approximator.fit(sa, target, **self.params["fit_params"])

Instances