def test_networkx():
"""Testing Graph object consistency for an edge-dictionary-type initialization object."""
try:
import networkx as nx
except ImportError:
return
ga = {0: {0: 1., 1: 1., 3: 3.},
1: {0: 1., 3: 2.},
2: {0: 2., 1: 3., 3: 1.},
3: {0: 1.}}
ganl = {0: 'l1', 1: 'l2', 2: 'l3', 3: 'l4'}
gael = {(0, 0): 'el1', (0, 1): 'el2', (0, 3): 'el3',
(1, 0): 'el4', (1, 3): 'el5', (2, 0): 'el6',
(2, 1): 'el7', (2, 3): 'el8', (3, 0): 'el9'}
g = nx.DiGraph()
for n in ganl.keys():
g.add_node(n, nl=ganl[n])
for e in gael.keys():
g.add_edge(e[0], e[1], w=ga[e[0]][e[1]], el=gael[e])
gs = list(graph_from_networkx([g], 'nl', 'el', 'w'))[0]
assert(gs[0] == ga and gs[1] == ganl and gs[2] == gael)
def graph_similarity(G, H, method='random_walk', **kwargs):
"""
Parameters
----------
G, H: nx.Graph
"""
assert method in ['random_walk']
if method == 'random_walk':
kernel = RandomWalk(**kwargs)
return kernel.fit_transform(graph_from_networkx([G, H]))[0, 1]
for n in range(3, 103):
Gs.append(nx.path_graph(n))
y.append(0)
Gs.append(nx.cycle_graph(n))
y.append(1)
############## Question 2
# Classify the synthetic graphs using graph kernels
# Split dataset into a training and a test set
# hint: use the train_test_split function of scikit-learn
G_train, G_test, y_train, y_test = train_test_split(Gs, y, test_size=0.1)
# Transform NetworkX graphs to objects that can be processed by GraKeL
G_train = list(graph_from_networkx(G_train))
G_test = list(graph_from_networkx(G_test))
# Use the shortest path kernel to generate the two kernel matrices ("K_train" and "K_test")
# hint: the graphs do not contain node labels. Set the with_labels argument of the the shortest path kernel to False
gk = ShortestPath(with_labels=False)
K_train = gk.fit_transform(G_train)
K_test = gk.transform(G_test)
clf = SVC(kernel='precomputed', C=1) # Initialize SVM
clf.fit(K_train, y_train) # Train SVM
y_pred = clf.predict(K_test) # Predict
# Compute the classification accuracy
def get_all(
self, output_type="dense", idxs=None, train_idxs=None,
shuffle=False, name_suffix=""):
ds_name = self.name + name_suffix
if shuffle:
if idxs is None:
idxs = np.arange(self.num_graphs)
else:
idxs = np.array(idxs)
np.random.shuffle(idxs)
if output_type == "dense":
graphs, targets = self.dataset
if idxs is not None:
graphs = graphs[idxs]
targets = targets[idxs]
ds = ds_utils.to_dense_ds(
graphs, targets, self._dim_node_features, self._num_node_labels)
ds = ds.batch(self.dense_batch_size)
elif output_type == "grakel":
graphs, targets = self.dataset
if idxs is not None:
graphs = graphs[idxs]
targets = targets[idxs]
if self._num_node_labels > 0:
node_labels_tag = "label"
elif self.node_one_labels:
node_labels_tag = "label_one"
else:
node_labels_tag = None
if self._num_edge_labels > 0:
edge_labels_tag = "label"
elif self.edge_one_labels:
edge_labels_tag = "label_one"
else:
edge_labels_tag = None
return gk.graph_from_networkx(
graphs,
node_labels_tag=node_labels_tag,
edge_labels_tag=edge_labels_tag), targets
# Custom kernel:
elif callable(output_type):
_, targets = self.dataset
gram = self._compute_gram_matrix(output_type)
if gram is None:
return
if idxs is not None:
train_idxs = idxs if train_idxs is None else train_idxs
gram, targets = gram[idxs, :][:, train_idxs], targets[idxs]
return gram, targets
elif output_type == "wl2":
batches = self._get_wl2_batches(ds_name, idxs)
if batches is None:
return
if isinstance(batches, tf.data.Dataset):
batches.name = ds_name
return batches
ds = ds_utils.wl_batches_to_dataset(*batches)
elif output_type == "wl2c":
batches = self._get_wl2c_batches(ds_name, idxs)
if batches is None:
return
if isinstance(batches, tf.data.Dataset):
batches.name = ds_name
return batches
ds = ds_utils.wl_batches_to_dataset(*batches, compact=True)
elif output_type == "en":
batches = self._get_en_batches(ds_name, idxs)
if batches is None:
return
if isinstance(batches, tf.data.Dataset):
batches.name = ds_name
return batches
ds = ds_utils.wl_batches_to_dataset(*batches, en_encode=True)
elif output_type == "wl1":
batches = self._get_wl1_batches(ds_name, idxs)
if batches is None:
return
if isinstance(batches, tf.data.Dataset):
batches.name = ds_name
return batches
ds = ds_utils.wl_batches_to_dataset(*batches, wl1_encode=True)
ds.name = ds_name
return ds
def run_samples_lasso(N, B, alpha, theta1, theta2, s1, s2):
import myKernels.RandomWalk as rw
test_info = pd.DataFrame()
k = theta1.shape[0]
for sample in tqdm.tqdm(range(N)):
Gs1 = []
Gs2 = []
error_1 = []
error_2 = []
n = 50
for i in range(50):
x1 = np.random.multivariate_normal(mean=np.zeros(k),
cov=theta1,
size=100)
A1 = np.corrcoef(x1.T)
if alpha == 0:
np.fill_diagonal(A1, 0)
A1[np.abs(A1) < 1e-5] = 0
else:
gl = graphical_lasso(A1, alpha=alpha, max_iter=1000)
A1 = gl[0]
A1[np.abs(A1) < 1e-5] = 0
np.fill_diagonal(A1, 0)
Gs1.append(nx.from_numpy_matrix(A1))
error_1.append(
np.sum(
np.logical_xor(
np.abs(np.triu(A1, 1)) > 0,
np.abs(np.triu(theta1, 1)) > 0)))
x2 = np.random.multivariate_normal(mean=np.zeros(k),
cov=theta2,
size=100)
A2 = np.corrcoef(x2.T)
if alpha == 0:
np.fill_diagonal(A2, 0)
A2[np.abs(A2) < 1e-5] = 0
else:
gl = graphical_lasso(A2, alpha=alpha, max_iter=1000)
A2 = gl[0]
A2[np.abs(A2) < 1e-5] = 0
np.fill_diagonal(A2, 0)
Gs2.append(nx.from_numpy_matrix(A2))
error_2.append(
np.sum(
np.logical_xor(
np.abs(np.triu(A2, 1)) > 0,
np.abs(np.triu(theta2, 1)) > 0)))
Gs = Gs1 + Gs2
try:
#rw_kernel = rw.RandomWalk(Gs, c = 0.0001, normalize=0)
#K = rw_kernel.fit_ARKU_plus(r = 6, normalize_adj=False, edge_attr= None, verbose=False)
graph_list = gk.graph_from_networkx(Gs)
kernel = [{"name": "SP", "with_labels": 0}]
init_kernel = gk.GraphKernel(kernel=kernel, normalize=0)
K = init_kernel.fit_transform(graph_list)
except:
continue
MMD_functions = [mg.MMD_b, mg.MMD_u]
kernel_hypothesis = mg.BoostrapMethods(MMD_functions)
function_arguments = [dict(n=n, m=n), dict(n=n, m=n)]
kernel_hypothesis.Bootstrap(K, function_arguments, B=B)
#print(f'p_value {kernel_hypothesis.p_values}')
#print(f"MMD_u {kernel_hypothesis.sample_test_statistic['MMD_u']}")
test_info = pd.concat(
(test_info,
pd.DataFrame(
{
'p_val': kernel_hypothesis.p_values['MMD_u'],
'sample': sample,
'mean_error_1': np.mean(error_1),
'mean_error_2': np.mean(error_2),
'alpha': alpha,
's1': s1,
's2': s2
},
index=[0])),
ignore_index=True)
return test_info
