import scipy
import numpy as np
from sklearn.cluster import KMeans
from utils.data_helper import get_laplacian
__all__ = ['spectral_clustering', 'get_L_cluster_cut']
def spectral_clustering(L, K, seed=1234):
"""
Implement paper "Shi, J. and Malik, J., 2000. Normalized cuts and image
segmentation. IEEE Transactions on pattern analysis and machine intelligence,
22(8), pp.888-905."
Args:
L: graph Laplacian, numpy or scipy matrix
K: int, number of clusters
Returns:
node_label: list
N.B.: for simplicity, we only consider simple and undirected graph
"""
num_nodes = L.shape[0]
assert (K < num_nodes - 1)
eig, eig_vec = scipy.sparse.linalg.eigsh(
L, k=K, which='LM', maxiter=num_nodes * 10000, tol=0, mode='normal')
kmeans = KMeans(n_clusters=K, random_state=seed).fit(eig_vec.real)
return kmeans.labels_
def get_L_cluster_cut(L, node_label):
adj = L - np.diag(np.diag(L))
adj[adj != 0] = 1.0
num_nodes = adj.shape[0]
idx_row, idx_col = np.meshgrid(range(num_nodes), range(num_nodes))
idx_row, idx_col = idx_row.flatten().astype(
np.int64), idx_col.flatten().astype(np.int64)
mask = (node_label[idx_row] == node_label[idx_col]).reshape(
num_nodes, num_nodes).astype(np.float)
adj_cluster = adj * mask
adj_cut = adj - adj_cluster
L_cut = get_laplacian(adj_cut, graph_laplacian_type='L4')
L_cluster = get_laplacian(adj_cluster, graph_laplacian_type='L4')
return L_cluster, L_cut
