gaussianmixture.py 文件源码

def _maximization(self):
        # Maximize priors
        priors = sum(self.responsibility_matrix)
        priors = [float(i)/sum(priors) for i in priors]

        # Maximize means
        mus = [0 for i in range(self.c)]
        for k in range(self.c):
            mus_k = sum(np.multiply(self.samples,
                                    self.responsibility_matrix[:, k][:, np.newaxis]))
            normalized_mus_k = mus_k / sum(self.responsibility_matrix[:, k])
            mus[k] = normalized_mus_k

        # Maximize covariances
        covs = [0 for i in range(self.c)]
        for k in range(self.c):
            covs[k] = np.cov(self.samples.T,
                             aweights=self.responsibility_matrix[:, k])

        return priors, mus, covs