snf.compute.group_predict

snf.compute.group_predict(train, test, labels, *, K=20, mu=0.4, t=20)[source]

Propagates labels from train data to test data via SNF

Parameters:
  • train (m-list of (S1, F) array_like) – Input subject x feature training data. Subjects in these data sets should have been previously labelled (see: labels).
  • test (m-list of (S2, F) array_like) – Input subject x feature testing data. These should be similar to the data in train (though the first dimension can differ). Labels will be propagated to these subjects.
  • labels ((S1,) array_like) – Cluster labels for S1 subjects in train data sets. These could have been obtained from some ground-truth labelling or via a previous iteration of SNF with only the train data (e.g., the output of sklearn.cluster.spectral_clustering() would be appropriate).
  • K ((0, N) int, optional) – Hyperparameter normalization factor for scaling. See Notes of snf.affinity_matrix for more details. Default: 20
  • mu ((0, 1) float, optional) – Hyperparameter normalization factor for scaling. See Notes of snf.affinity_matrix for more details. Default: 0.5
  • t (int, optional) – Number of iterations to perform information swapping during SNF. Default: 20
Returns:

predicted_labels – Cluster labels for subjects in test assigning to groups in labels
Return type:

(S2,) np.ndarray