distributions.py 文件源码

def _posterior_hypparams(self,n,tot,normalizers,feasible):
        if n == 0:
            return n, self.alpha_0, self.r_probs, self.r_support
        else:
            r_probs = self.r_probs[feasible]
            r_support = self.r_support[feasible]
            log_marg_likelihoods = special.betaln(self.alpha_0 + tot - n*r_support,
                                                        self.beta_0 + r_support*n) \
                                    - special.betaln(self.alpha_0, self.beta_0) \
                                    + normalizers
            log_marg_probs = np.log(r_probs) + log_marg_likelihoods
            log_marg_probs -= log_marg_probs.max()
            marg_probs = np.exp(log_marg_probs)

            return n, self.alpha_0 + tot, marg_probs, r_support