def __init__(self,
nodes_degree,
size=2,
down_sampling=0,
seed=1,
path_labels='data/',
input_file=None,
f_type=t.FloatTensor):
'''
:param nodes_degree: Dict with node_id: degree of node
:param size: projection space
:param downsampling: perform downsampling of common node
:param seed: seed for random function
:param index2node: index between a node and its representation
:param path_labels: location of the file containing the ground true (label for each node)
:param input_file: name of the file containing the ground true (label for each node)
:return:
'''
global float_tensor
float_tensor = f_type
self.down_sampling = down_sampling
self.seed = seed
if size % 4 != 0:
log.warn(
"consider setting layer size to a multiple of 4 for greater performance"
)
self.layer1_size = int(size)
if nodes_degree is not None:
self.build_vocab_(nodes_degree)
self.ground_true, self.k = load_ground_true(path=path_labels,
file_name=input_file)
# inizialize node and context embeddings
self.reset_weights()
self.compute_negative_sampling_weight()
else:
log.info("Model not initialized")
def __init__(self,
nodes_degree=None,
size=2,
down_sampling=0,
seed=1,
table_size=100000000,
path_labels='data/',
input_file=None):
'''
:param nodes_degree: Dict with node_id: degree of node
:param size: projection space
:param down_sampling: perform down_sampling of common node
:param seed: seed for random function
:param table_size: size of the negative table to generate
:param path_labels: location of the file containing the ground true (label for each node)
:param input_file: name of the file containing the ground true (label for each node)
:return:
'''
self.down_sampling = down_sampling
self.seed = seed
self.table_size = table_size
if size % 4 != 0:
log.warning(
"consider setting layer size to a multiple of 4 for greater performance"
)
self.layer1_size = int(size)
if nodes_degree is not None:
self.build_vocab_(nodes_degree)
self.ground_true, self.k = load_ground_true(path=path_labels,
file_name=input_file)
# initialize node and context embeddings
self.make_table()
self.precalc_sampling()
self.reset_weights()
else:
log.warning("Model not initialized, need the nodes degree")
f"./plots/{output_file}/animation_{come_model_type}_d{representation_size}_k{k}.gif",
writer='imagemagick')
# ### write predictions to labels_pred.txt
# save com_learner.g_mixture to file
joblib.dump(
com_learner.g_mixture,
f'./model/g_mixture_{output_file}_{come_model_type}_d{representation_size}_k{k}.joblib'
)
# using predictions from com_learner.g_mixture with node_embeddings
np.savetxt(
f'./data/{output_file}/labels_pred_{come_model_type}_d{representation_size}_k{k}.txt',
model.classify_nodes())
# ### NMI
labels_true, _ = load_ground_true(path="data/" + input_file,
file_name=input_file)
print("labels_true: ", labels_true)
if labels_true is not None:
nmi = metrics.normalized_mutual_info_score(labels_true,
node_classification)
print(
f"===NMI=== for type={come_model_type} with d={representation_size} and K={k}: ",
nmi)
else:
print(
f"===NMI=== for type={come_model_type} with d={representation_size} and K={k} could not be computed"
)
# ### plotting
plot_name = f"{come_model_type}_d{representation_size}_k{k}"
if should_plot: