def reward(self, s=None, a=None, ns=None):
t_dist = self.transition_reward_dist(s, a)
r_dist = {}
for (ns_, r), p in t_dist.items():
if ns_ == ns:
r_dist[r] = r_dist.get(r, 0) + p
return sample_prob_dict(r_dist)
def act(self, s, softmax_temp=None, randchoose=None):
actions = self.ham.available_actions(s)
a_q = {}
for a in actions:
q = self._qval(s, a)
# if q == -np.inf: #never take actions that machine loop
# continue
a_q[a] = q
if softmax_temp is None:
softmax_temp = self.softmax_temp
if randchoose is None:
randchoose = self.randchoose
adist = calc_esoftmax_dist(a_q,
temp=softmax_temp,
randchoose=randchoose)
return sample_prob_dict(adist)
def transition(self, s, a):
return sample_prob_dict(self.transition_reward_dist(s, a))[0]
def reward(self, s=None, a=None, ns=None):
return sample_prob_dict(self.reward_dist(s, a, ns))
def observation(self, a: Action, ns: State) -> Observation:
return sample_prob_dict(self.observation_dist(a, ns))
def get_init_state(self):
return sample_prob_dict(self.get_init_state_dist())
def act(self, s, softmax_temp=None, randchoose=None):
adist = self.act_dist(s, softmax_temp, randchoose)
return sample_prob_dict(adist)
def transition_reward(self, s=None, a=None):
return sample_prob_dict(self.transition_reward_dist(s, a))
def get_init_state(self):
if len(self.init_state_dist) == 0:
raise ValueError("No initial state defined")
return sample_prob_dict(self.init_state_dist)
def transition(self,
s: "state",
ja: "joint action"):
return sample_prob_dict(self.transition_reward_dist(s, ja))[0]