else:
node_num = self._node_num
adj_mtx_r = np.zeros((node_num, node_num))
for v_i in range(node_num):
for v_j in range(node_num):
if v_i == v_j:
continue
adj_mtx_r[v_i, v_j] = self.get_edge_weight(v_i, v_j)
return adj_mtx_r
if __name__ == '__main__':
# load Zachary's Karate graph
edge_f = 'data/karate.edgelist'
G = graph_util.loadGraphFromEdgeListTxt(edge_f, directed=False)
G = G.to_directed()
res_pre = 'results/testKarate'
graph_util.print_graph_stats(G)
t1 = time()
embedding = HOPE(4, 0.01)
embedding.learn_embedding(graph=G,
edge_f=None,
is_weighted=True,
no_python=True)
print('HOPE:\n\tTraining time: %f' % (time() - t1))
viz.plot_embedding2D(embedding.get_embedding()[:, :2],
di_graph=G,
node_colors=None)
plt.show()
def call_exps(params, data_set):
# Load Dataset
print('Dataset: %s' % data_set)
######## for SBM, r_mat, hyperbolic
#if data_set[10:13] == 'r_m' or data_set[10:13] == 'sto' or data_set[10:13] == 'hyp':
# di_graph = nx.read_gpickle('gem/data/%s/graph.gpickle' % data_set)[0]
#else:
#di_graph = nx.read_gpickle('gem/data/%s/graph.gpickle' % data_set)[0]
di_graph = nx.read_gpickle('gem/data/%s/graph.gpickle' % data_set)
di_graph, nodeListMap = graph_util.get_lcc(di_graph)
try:
os.makedirs('gem/nodeListMap')
except:
pass
pickle.dump(nodeListMap, open('gem/nodeListMap/%s.pickle' % data_set,
'wb'))
graph_util.print_graph_stats(di_graph)
# Load node labels if given
if bool(params["node_labels"]):
node_labels = pickle.load(
open('gem/data/%s/node_labels.pickle' % data_set, 'rb'))
node_labels_gc = np.zeros(
(di_graph.number_of_nodes(), node_labels.shape[1]))
for k, v in nodeListMap.items():
try:
node_labels_gc[v, :] = node_labels[k, :].toarray()
# Already a numpy array
except AttributeError:
node_labels_gc[v, :] = node_labels[k, :]
node_labels = node_labels_gc
else:
node_labels = None
# Search through the hyperparameter space
if params["find_hyp"]:
choose_best_hyp(data_set, di_graph, node_labels, params)
# Load best hyperparameter and test it again on new test data
for d, meth, exp in itertools.product(params["dimensions"],
params["methods"],
params["experiments"]):
dim = int(d)
MethClass = getattr(importlib.import_module("gem.embedding.%s" % meth),
methClassMap[meth])
opt_hyp_f_pre = 'gem/experiments/config/synthetic/%s_%s_%s' % (
data_set, meth, params["samp_scheme"])
try:
if exp != "viz":
if exp == 'lpt':
model_hyp = json.load(
open('%s_lp.conf' % opt_hyp_f_pre, 'r'))
else:
model_hyp = json.load(
open('%s_%s.conf' % (opt_hyp_f_pre, exp), 'r'))
else:
model_hyp = json.load(
open('%s_%s.conf' % (opt_hyp_f_pre, params["viz_params"]),
'r'))
except IOError:
print('Default hyperparameter of the method chosen')
model_hyp = json.load(
open('gem/experiments/config/%s.conf' % meth, 'r'))
hyp = {}
hyp.update(model_hyp[meth])
hyp.update({"d": dim})
if meth == "sdne":
hyp.update({
"modelfile": [
"gem/intermediate/en_mdl_%s_%d.json" % (data_set, dim),
"gem/intermediate/dec_mdl_%s_%d.json" % (data_set, dim)
],
"weightfile": [
"gem/intermediate/enc_wts_%s_%d.hdf5" % (data_set, dim),
"gem/intermediate/dec_wts_%s_%d.hdf5" % (data_set, dim)
]
})
elif meth == "gf" or meth == "node2vec":
hyp.update({"data_set": data_set})
MethObj = MethClass(hyp)
run_exps(MethObj, meth, dim, di_graph, data_set, node_labels, params)