"""
Settings : default -> for Kipf GCN settings, quick -> for running the whole thing fast (to check that everything works)
labels_percent_list : determines how many labeled nodes will be used for training. Percent with respect to the training set, 100% means the whole training set
list adj : used to display stats on connection to known nodes of wrongly/correctly classified nodes.
maintain_label_balance_list : will (try) to keep the same ratio of classes in labeled training set
with_test_features_list : Hides testing features by modifying the Adj matrix if set to False
models_list : Models to try. Subsample GCN strictly uses subsmaple nodes, GCN uses all the features
"""
if __name__ == "__main__":
# Tensorflow settings
flags = tf.app.flags
FLAGS = flags.FLAGS
settings = graph_settings()['default']
set_tf_flags(settings['params'], flags, verbose=True)
# Verbose settings
SHOW_TEST_VAL_DATASET_STATS = False
VERBOSE_TRAINING = False
# Random seed
seed = settings['seed']
np.random.seed(seed)
SAMPLING_TRIALS = 16 # How many time the same experiment will be repeated to get standard dev.
# Load data. Features and labels.
adj, features, y_train, y_val, y_test, initial_train_mask, val_mask, test_mask = load_data(
FLAGS.dataset)
# Some preprocessing
features = preprocess_features(features)
'incorrect_paths_to_known': incorrect_paths_to_known
}
}
pk.dump(dict_output,
open(
os.path.join('results', fileinfo + 'w_test_features=' + str(WITH_TEST) + '_label_balance=' +
str(MAINTAIN_LABEL_BALANCE) + '_results.p'), 'wb'))
result.clear()
if __name__ == "__main__":
# Settings
flags = tf.app.flags
FLAGS = flags.FLAGS
settings = graph_settings()['default']
set_tf_flags(settings['params'], flags)
# Verbose settings
SHOW_TEST_VAL_DATASET_STATS = False
VERBOSE_TRAINING = False
# Random seed
seed = settings['seed']
np.random.seed(seed)
# Load data
adj, features, y_train, y_val, y_test, initial_train_mask, val_mask, test_mask = load_data(FLAGS.dataset)
# Some preprocessing
features = preprocess_features(features)
labels_percent_list = [0.1, 5, 10, 20, 40, 50, 75, 100]
list_adj = get_adj_powers(adj.toarray())