def update(self, arm, reward):
"""更新收益
:param reward: 收益
:type arm: 选中的臂的下标
"""
Policy.update(self, arm, reward)
self.b[arm] = self.b[arm] + reward * self.context[arm]
self.context[arm].shape = (self.d, 1)
self.A[arm] = self.A[arm] + self.context[arm].dot(np.transpose(self.context[arm]))
def update(self, arm, reward):
Policy.update(self, arm, reward)
ratio = math.exp(self.gamma * reward /
(self.n_bandits * self._probs[arm]))
# todo 权重会越来越大？
self._weights[arm] *= ratio