multidim.PointCloud.from_multisample_multilabel

classmethod PointCloud.from_multisample_multilabel(list_of_samples, list_of_labels, equal_priors=True, normalize_domain=False)

Produce a single labeled and weighted pointcloud from a list of samples and labels of those samples.

Parameters:

list_of_samples :

A list (or np array) of np arrays. Each such array is considered to be a sample of N points in R^d. N can vary between entries, but d cannot.

list_of_labels :

A list of labels. Labels can be anything, but it is covenient to use strings like “red” and “blue”. list_of_labels[i] is the label for the points in list_of_samples[i].

equal_priors:

Re-normalize weights so that each label is equally likely. Default: True

normalize_domain:

Use SVD/PCA to re-shape the original data to be roughy spherical. This should allow better learning via CDER. Default: False