skinning_clusters

skinning_clusters(W, D, T, k, l=2, num_clustering_features=10, return_centroids=False, return_simplex_features=False)

Skinning clusters.

Parameters:

Name	Type	Description	Default
W	numpy float array	n x b skinning weights	required
D	numpy float array	b x 1 eigenvalue/weighing given to each skinning weight	required
T	numpy int array	T x 4 tet geometry	required
k	int	number of clusters	required
l	int	power to raise D to	2
num_clustering_features	int	number of features to use for clustering	10
return_centroids	bool	whether to return the centroids of the clusters	False
return_simplex_features	bool	whether to return the features averaged over each tet	False

Source code in src\fast_cody\skinning_clusters.py

def skinning_clusters(W, D, T, k, l=2, num_clustering_features=10,
                      return_centroids=False, return_simplex_features=False):
    """ Skinning clusters.

    Parameters
    ----------
    W : numpy float array
        n x b skinning weights
    D : numpy float array
        b x 1 eigenvalue/weighing given to each skinning weight
    T : numpy int array
        T x 4 tet geometry
    k : int
        number of clusters
    l : int
        power to raise D to
    num_clustering_features : int
        number of features to use for clustering
    return_centroids : bool
        whether to return the centroids of the clusters
    return_simplex_features : bool
        whether to return the features averaged over each tet
    """
    num_clustering_features = min(num_clustering_features, W.shape[1])
    # need to average the skinning weights over each tet
    assert(T.shape[1] == 4, "only tets implemented so far for clustering")

    Wt = average_onto_simplex(W, T)
    # Wt2 = Wt / np.power(D, 2)
    Wt = Wt / np.power(D, l)
    Wt = Wt[:, 0:num_clustering_features]
    kmeans = KMeans(n_clusters=k, random_state=0).fit(Wt)
    l = kmeans.labels_

    if return_simplex_features:
        return l, Wt
    if return_centroids == True:
        return l, kmeans.cluster_centers_
    else:
        return l