axis=1)[:, None]
# Number of examples
examples_x_class = np.ceil(examples_x_class *
np.array(args.division)[:, None])
else:
# Balanced distribution of nodes
examples_x_class = np.tile(
np.ceil(np.array(args.division) / num_class)[:, None],
(1, num_class))
# Iterate class prototypes
for i in range(num_class):
el = Element(proto_files[i])
# Define distortions
el.set_distortion(args.nodeDisplace, args.nodeAdd, args.edgeMaximum,
args.addEdge, args.rmEdge, args.edgeConnection)
# Add nodes if necessary
if args.nodeThreshold is not None:
el.add_nodes(args.nodeThreshold)
# For each set (train, validation, test)
for s in range(len(f_set)):
# Apply random distortion to generate each element
for sample in range(int(examples_x_class[s][i])):
g = el.distort()
nx.write_gml(
g, args.dirDataset + f_set[s] + str(sample) + '_' +
el.get_label() + '.gml')
