3f2299f9fe6b8928f1c8c576203d6bf5c4b28758,pymc3/variational/stein.py,Stein,dlogp,#Stein#,26

Before Change

        )
        dlogp = tt.concatenate(list(map(unpack, gradients_for_rmatrices)), axis=-1)

        if self.use_histogram:
            dlogp = theano.clone(
                dlogp,
                dict(zip(self.approx.symbolic_randoms, self.approx.collect("histogram")))
            )
        return dlogp

    @node_property
    def grad(self):

After Change

    @property
    def dlogp(self):
        return tt.grad(
            self.logp_norm.sum(),
            self.input_matrix
        )

    @node_property
    def grad(self):
        n = floatX(self.input_matrix.shape[0])

Instances