def p_r__aseq(self, a_seq, beta=1): # P(r | a)
pos = self.agent
p_move__item = np.ones((1, len(self.items)))
for a in a_seq:
p_move__item *= self._p_move__item(pos, beta)[[a]]
pos = pos + self.move_action.actions[a]
prob = np.dot(p_move__item, self.p_item__reward)
return probutil.normalized_2d_array(prob, 1)[0]
def p_dist_a__r(self, s=None, beta=1): # P(r | a)
s = self.agent if s is None else s
p_move__item = self._p_move__item(s, beta)
prob = np.dot(p_move__item, self.p_item__reward)
return probutil.normalized_2d_array(prob, 0)
def p_aseq__r(self, beta=1): # P(r | a)
p_move__item = self._p_move__item(self.agent, beta)
prob = np.dot(p_move__item, self.p_item__reward)
return probutil.normalized_2d_array(prob, 1)
def p_r__a(self, beta=1): # P(r | a)
return probutil.normalized_2d_array(self.p_item__reward, 1)