Refactoring

[FarnazH]

# cluster data
# X = features
clustering = DBSCAN(eps=3, min_samples=2).fit(X)
labels = clustering.labels_

# as a class
method = ClusteringSelection()
method.select(labels, num_select)  # return selected indices

# as a function
select_from_clusters(num_select, labels, selection_algorithm=OptiSim(r, k))

# compute distance

# arr is treated as a feature matrix
selected = OptiSim(r, k).select(arr, num_select, func_distance='euclidean', labels=None)
# arr is treated as a distance matrix because func_distance is None
selected = OptiSim(r, k).select(arr, num_select, func_distance=None, labels=None)

# arr is treated as a distance matrix
selected = OptiSim(r, k, compute_distance=False).select(arr, num_select, labels=None)

selected = GridPartitioning(...).select(arr, num_select, labels=None)