d1ac7b831ad36cd0e4bdd7980819f83208345148,gpflow/expectations.py,,_expectation,#Any#Any#Any#Any#Any#,633

Before Change

            Xmu = tf.identity(Xmu)

        N = tf.shape(Xmu)[0] - 1
        Xmum = Xmu[:-1, :]
        Xmup = Xmu[1:, :]
        op = tf.expand_dims(Xmum, 2) * tf.expand_dims(Xmup, 1) + Xcov[1, :-1, :, :]  // NxDxD
        return kern.variance * tf.matmul(tf.tile(tf.expand_dims(Z, 0), (N, 1, 1)), op)

After Change

        N = tf.shape(Xmu)[0] - 1
        var_Z = kern.variance * feat.Z  // MxD
        tiled_Z = tf.tile(tf.expand_dims(var_Z, 0), (N, 1, 1))  // NxMxD
        eXX = Xcov[1, :-1] + (Xmu[:-1][..., None] * Xmu[1:][:, None, :])  // NxDxD
        return tf.matmul(tiled_Z, eXX)


@dispatch((Gaussian, DiagonalGaussian), kernels.Linear, InducingPoints, kernels.Linear, InducingPoints)
def _expectation(p, kern1, feat1, kern2, feat2):

Instances