Method predict

rl/approx_control.py:69–72

(self, s, a)

    self.w = np.zeros(dims)

  def predict(self, s, a):
    sa = merge_state_action(s, a)
    x = self.featurizer.transform([sa])[0]
    return x @ self.w

  def predict_all_actions(self, s):
    return [self.predict(s, a) for a in ALL_POSSIBLE_ACTIONS]