def step(self, X):
pdist_na = self.f_probs(X)
acts_n = cat_sample(pdist_na)
return {
"action" : acts_n,
"pdist" : pdist_na
}
def step(self, s_n):
f_na = self.f_sa[s_n]
prob_nk = softmax_prob(f_na)
acts_n = cat_sample(prob_nk)
return {"action":acts_n,
"pdist" : f_na}
def step(self, X):
pdist_na = self.f_probs(X)
acts_n = cat_sample(pdist_na)
return {"action": acts_n, "pdist": pdist_na}