lp_solutions_matrix {metasnf} | R Documentation |
Label propagate cluster solutions to unclustered subjects
Description
Given a solutions_matrix derived from training subjects and a full_data_list containing both training and test subjects, re-run SNF to generate a total affinity matrix of both train and subjects and use the label propagation algorithm to assigned predicted clusters to test subjects.
Usage
lp_solutions_matrix(
train_solutions_matrix,
full_data_list,
distance_metrics_list = NULL,
weights_matrix = NULL,
verbose = FALSE
)
Arguments
train_solutions_matrix |
A solutions_matrix derived from the training set. The propagation algorithm is slow and should be used for validating a top or top few meaningful chosen clustering solutions. It is advisable to use only a small subset of rows from the original training solutions_matrix for label propagation. |
full_data_list |
A data_list containing subjects from both the training and testing sets. |
distance_metrics_list |
Like above - the distance_metrics_list (if any) that was used for the original batch_snf call. |
weights_matrix |
Like above. |
verbose |
If TRUE, print progress to console. |
Value
labeled_df a dataframe containing a column for subjectkeys, a column for whether the subject was in the train (original) or test (held out) set, and one column per row of the solutions matrix indicating the original and propagated clusters.