a1dd11c68e5911f069a747c917e2a4bfdd5ae4f4,geomstats/learning/em_expectation_maximization.py,RiemannianEM,update_variances,#RiemannianEM#Any#Any#Any#,131

Before Change

                                                                                                           g_index].sum()
                self.variances[:, g_index] = (self.normalization_factor.phi(dtm)).data.numpy()
            else:
                dtm = ((distance(z.unsqueeze(1).expand(N, M, D),
                                 self.means.unsqueeze(0).expand(N, M, D)) ** 2) * wik).sum(0) / wik.sum(0)
                // print("dtms ", dtm.size())
                self.variances = (self.normalization_factor.phi(dtm)).data.numpy()

    def _expectation(self, data):

After Change

        else:


            z_gs = gs.expand_dims(z.data.numpy(), 1)
            z_gs = gs.repeat(z_gs,M,axis = 1)
            means_gs = gs.expand_dims(self.means.data.numpy(),0)
            means_gs = gs.repeat(means_gs,N,axis = 0)

            z_torch = z.unsqueeze(1).expand(N, M, D)
            means_torch = self.means.unsqueeze(0).expand(N, M, D)
            wik_gs = wik.data.numpy()
            dtm_gs = ((self.riemannian_metric.dist(z_gs,
                             means_gs) ** 2) * wik_gs).sum(0) / wik_gs.sum(0)

            // dtm = ((distance(z.unsqueeze(1).expand(N, M, D),
            //                  self.means.unsqueeze(0).expand(N, M, D)) ** 2) * wik).sum(0) / wik.sum(0)

Instances