#reading the number of vertices,edges and partitions required
with open(f_name) as f:
first_line = f.readline()
graph_key = [int(s) for s in first_line.split() if s.isdigit()]
n_vert = graph_key[0]
n_edge = graph_key[1]
n_partition = graph_key[2]
#Constructing graph
G = nx.Graph()
G.add_nodes_from(range(n_vert))
G.add_edges_from(graph)
#Computing laplacian
L = nx.laplacian_matrix(G)
L = csr_matrix.astype(L, dtype='f')
# Computing eigen vectors
eig_val, eig_vec = la.eigsh(L, 2, which='SM')
def distance(p1, p2):
return np.sum((p1 - p2)**2)
# initialisation algorithm K++
def initialize(data, k):
'''
intialized the centroids for K-means++
inputs:
data - numpy array of data points having shape (200, 2)
labels1 = np.ndarray(shape=([len(ntypes1),2]),dtype=int)
labels2 = np.ndarray(shape=([len(ntypes2),2]),dtype=int)
for i, j in enumerate(sorted(ntypes1.keys())):
if j in zgats: labels1[i] = [1,0]
else: labels1[i] = [0,1]
for i, j in enumerate(sorted(ntypes2.keys())):
if j in zgats: labels2[i] = [1,0]
else: labels2[i] = [0,1]
feat1 = csr_matrix(features1)
feat2 = csr_matrix(features2)
feat_csr1 = csr_matrix.astype(feat1,np.float32)
feat_csr2 = csr_matrix.astype(feat2,np.float32)
x = open("ind.logdec.x","wt")
tx = open("ind.logdec.tx","wt")
allx = open("ind.logdec.allx","wt")
graf = open("ind.logdec.graph","wt")
y = open("ind.logdec.y","wt")
ty = open("ind.logdec.ty","wt")
ally = open("ind.logdec.ally","wt")
pkl.dump(labels2,y)
pkl.dump(labels2,ally)
pkl.dump(labels1,ty)
pkl.dump(feat_csr2,x)
pkl.dump(feat_csr2,allx)
def to_ndarray(l, dtype=int):
assert len(l) > 0
assert len(l[0]) > 0
rows = len(l)
cols = len(l[0])
arr = np.ndarray(shape=(rows, cols), dtype=int)
for i, row in enumerate(l):
arr[i] = row
return arr
labels_test = to_ndarray(labels_test)
labels_train = to_ndarray(labels_train)
labels_train_all = to_ndarray(labels_train_all)
feat_csr_test = csr_matrix.astype(csr_matrix(features_test), np.float32)
feat_csr_train = csr_matrix.astype(csr_matrix(features_train), np.float32)
feat_csr_train_all = csr_matrix.astype(csr_matrix(features_train_all),
np.float32)
f_x = open("ind.logdec.x", "wb")
f_tx = open("ind.logdec.tx", "wb")
f_allx = open("ind.logdec.allx", "wb")
f_graf = open("ind.logdec.graph", "wb")
f_y = open("ind.logdec.y", "wb")
f_ty = open("ind.logdec.ty", "wb")
f_ally = open("ind.logdec.ally", "wb")
print("label lengths (test, train, all, cnt):", len(labels_test),
len(labels_train), len(labels_train_all), len(nfanins)),
print("features size (test, train, all, cnt):", feat_csr_test.shape,
