Postprocessing module

Postprocessing module.

MARBLE.postprocessing.cluster(data, cluster_typ='kmeans', n_clusters=15, seed=0)

Cluster data.

MARBLE.postprocessing.distribution_distances(data, cluster_typ='kmeans', n_clusters=None, seed=0)

Return distance between datasets.

Returns:

PyG data object containing .out attribute, a nx2 matrix of embedded data clusters: sklearn cluster object dist (cxc matrix): pairwise distances where c is the number of clusters

Return type:

data

MARBLE.postprocessing.embed_in_2D(data, embed_typ='umap', manifold=None, seed=0)

Embed into 2D via for visualisation.

Parameters:

• data – PyG input data

• embed_typl (string, optional) – Embedding algorithm to use (tsne, umap, PCA)

• manifold (sklearn object, optional) – Manifold object returned by some embedding algorithms (PCA, umap). Useful when trying to compare datasets.

• seed (int, optional) – Random seed. The default is 0.

Returns:

PyG data object containing emb_2D attribute.