ac8218b204a2c8a972f8f2410a3cc150a4ef3afc,geomstats/learning/frechet_mean.py,,variance,#Any#Any#Any#Any#Any#,19

Before Change

    weights : array-like, shape=[n_samples, 1], optional
    
    n_points = gs.shape(points)[0]
    if weights is None:
        weights = gs.ones((n_points,))
    weights = gs.array(weights)

    sum_weights = gs.sum(weights)
    if point_type == "vector":
        points = gs.to_ndarray(points, to_ndim=2)
        base_point = gs.to_ndarray(base_point, to_ndim=2)
        // weights = gs.to_ndarray(weights, to_ndim=2, axis=1)
    if point_type == "matrix":
        points = gs.to_ndarray(points, to_ndim=3)
        base_point = gs.to_ndarray(base_point, to_ndim=3)
        // weights = gs.to_ndarray(weights, to_ndim=3, axis=1)
        // weights = weights[:, :, 0]

    sq_dists = metric.squared_dist(base_point, points)
    var = gs.einsum("...,...->...", weights, sq_dists)

    var = gs.sum(var)
    var /= sum_weights

    if n_points == 1 and gs.ndim(var) > 0:
        var = gs.squeeze(var, axis=0)
    return var


def linear_mean(points, weights=None):

After Change

    weights : array-like, shape=[n_samples, 1], optional
    
    n_points = 1
    if point_type == "vector" and gs.ndim(points) == 2:
        n_points = gs.shape(points)[0]
    if point_type == "matrix" and gs.ndim(points) == 3:
        n_points = gs.shape(points)[0]

    if weights is None:
        weights = gs.ones((n_points,))

    sum_weights = gs.sum(weights)
    sq_dists = metric.squared_dist(base_point, points)
    var = weights * sq_dists

    var = gs.sum(var)
    var /= sum_weights

Instances