mescla.emma.endmembers.rank_candidates

mescla.emma.endmembers.rank_candidates(U_samples, U_candidates, subspace_distances=None, labels=None)[source]

Score candidate end-members for extremeness and subspace membership.

Parameters:

  • U_samples (ndarray, shape (n_samples, n_components)) – Mixed samples in U-space.

  • U_candidates (ndarray, shape (n_candidates, n_components)) – Candidates projected with the same standardiser and eigenvectors.

  • subspace_distances (ndarray, optional) – From mescla.emma.projection.subspace_distance(). Small is good.

  • labels (tuple of str, optional)

Returns:

DataFrame – extremeness (distance from the sample centroid in U-space, larger is better – an end-member must be more extreme than the mixtures it explains) and subspace_distance (smaller is better).

Return type:

pd.DataFrame