'node2vec_emb_folder':
f"./saved_data/node2vec_emb/{dataset_name}_stability",
'sdne_emb_folder': f"./saved_data/sdne_emb/{dataset_name}_stability",
}
params = SettingParam(**params)
# params = read_config_file(filepath="./stability_configuration.ini", config_task="stability")
is_reconstruction_mAP = True
k_query = [100, 200, 500, 1000, 2000, 4000, 6000, 8000, 10000]
# ==================== Data =========================
graphs, idx2node = read_dynamic_graph(folder_path=params.dataset_folder,
limit=None,
convert_to_idx=True)
print("Origin graphs:")
for i, g in enumerate(graphs):
print_graph_stats(g, i, end="\t")
print(f"Isolate nodes: {nx.number_of_isolates(g)}")
# draw_graph(g, limit_node=25)
# ========= DynGEM ===========
if params.is_dyge:
print("=============== DynGEM ============")
# -------- Training ----------
dy_embeddings = load_dy_embeddings(params.dyge_emb_folder)
if is_reconstruction_mAP:
for i, g in enumerate(graphs):
reconstruction_prec = reconstruction_precision_k(
embedding=dy_embeddings[i], graph=g, k_query=k_query)
print(reconstruction_prec)
# ============== Node2Vec ============
def save_embedding(self, filepath):
if self.embedding is None:
self.embedding = self.get_embedding()
with open(filepath, 'wb') as fp:
pickle.dump(self.embedding, fp)
def load_embedding(self, filepath):
with open(filepath, 'rb') as fp:
self.embedding = pickle.load(fp)
return self.embedding
if __name__ == "__main__":
# G = get_graph_from_file(filename="../data/email-eu/email-Eu-core.txt")
G = nx.gnm_random_graph(n=40, m=100, seed=6)
print_graph_stats(G)
pos = nx.spring_layout(G, seed=6)
draw_graph(G, limit_node=50, pos=pos)
# print(G.edges)
embedding_dim = 4
g_hidden_df, hidden_G = preprocessing_graph_for_link_prediction(
G=G, drop_node_percent=0.2, edge_rate=0.1)
ge = TStaticGE(G=hidden_G,
embedding_dim=embedding_dim,
hidden_dims=[16, 8],
l2=1e-5,
alpha=0.2,
beta=10,