Skip to content

utils

estimate_spacing(coords)

Estimate the spacing between points in a point cloud. This is just the median distance between nearest neighbors.

Parameters:

Name Type Description Default
coords NDArray[floating]

The point cloud to estimate spacing of. Should have shape (npoint, ndim).

 required
Source code in megham/utils.py 
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
def estimate_spacing(coords: NDArray[np.floating]):
    """
    Estimate the spacing between points in a point cloud.
    This is just the median distance between nearest neighbors.

    Parameters
    ----------
    coords: NDArray[np.floating]
        The point cloud to estimate spacing of.
        Should have shape (npoint, ndim).
    """
    edm = make_edm(coords)
    edm[edm == 0] = np.nan
    nearest_dists = np.nanmin(edm, axis=0)

    return np.median(nearest_dists)

estimate_var(src, dst, dim_groups=None)

Estimate variance between point clouds for use with something like a GMM.

Parameters:

Name Type Description Default
src NDArray[floating]

The set of source points to be mapped onto the target points. Should have shape (nsrcpoints, ndim).

 required
dst NDArray[floating]

The set of destination points to be mapped onto. Should have shape (ndstpoints, ndim).

 required
dim_groups Optional[Sequence[Sequence[int] | NDArray[int_]]]

Which dimensions should be computed together. If None all dimensions will be treated seperately.

 None

Returns:

Name Type Description
var NDArray[floating]

The estimated variance. Will have shape (ndim,).
Source code in megham/utils.py 
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
def estimate_var(
    src: NDArray[np.floating],
    dst: NDArray[np.floating],
    dim_groups: Optional[Sequence[Sequence[int] | NDArray[np.int_]]] = None,
) -> NDArray[np.floating]:
    """
    Estimate variance between point clouds for use with something like a GMM.

    Parameters
    ----------
    src : NDArray[np.floating]
        The set of source points to be mapped onto the target points.
        Should have shape (nsrcpoints, ndim).
    dst : NDArray[np.floating]
        The set of destination points to be mapped onto.
        Should have shape (ndstpoints, ndim).
    dim_groups : Optional[Sequence[Sequence[int] | NDArray[np.int_]]], default: None
        Which dimensions should be computed together.
        If None all dimensions will be treated seperately.

    Returns
    -------
    var : NDArray[np.floating]
        The estimated variance.
        Will have shape (ndim,).
    """
    nsrcpoints, ndim = src.shape
    ndstpoints = len(dst)

    if dim_groups is None:
        dim_groups = [[dim] for dim in range(ndim)]
    else:
        dims_flat = np.concatenate(dim_groups)
        no_group = np.setdiff1d(np.arange(ndim), dims_flat)
        dim_groups = list(dim_groups)
        dim_groups = dim_groups + [[dim] for dim in no_group]

    var = np.zeros(ndim)
    for dim_group in dim_groups:
        sq_diff = dist.cdist(src[:, dim_group], dst[:, dim_group], metric="sqeuclidean")
        var[dim_group] = np.nansum(sq_diff) / (len(dim_group) * nsrcpoints * ndstpoints)

    return var

make_edm(coords)

Make an Euclidean distance matrix from a set of points.

Parameters:

Name Type Description Default
coords NDArray[floating]

The (npoint, ndim) array of input points.

 required

Returns:

Name Type Description
edm NDArray[floating]

The (npoint, npoint) euclidean distance matrix.
Source code in megham/utils.py 
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
def make_edm(coords: NDArray[np.floating]) -> NDArray[np.floating]:
    """
    Make an Euclidean distance matrix from a set of points.

    Parameters
    ----------
    coords : NDArray[np.floating]
        The (npoint, ndim) array of input points.

    Returns
    -------
    edm : NDArray[np.floating]
        The (npoint, npoint) euclidean distance matrix.
    """
    dist_vec = dist.pdist(coords)
    edm = dist.squareform(dist_vec)

    return edm