Method predict

rl/cartpole.py:56–59

(self, s, a)

    self.w = np.zeros(dims)

  def predict(self, s, a):
    sa = np.concatenate((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 range(self.env.action_space.n)]