parser.add_argument('--num_hidden', type=int, default=64)
parser.add_argument('--num_layers', type=int, default=8)
parser.add_argument('--dropout', type=float, default=0.5)
parser.add_argument('--alpha', type=float, default=0.1)
parser.add_argument('--lamda', type=float, default=0.5)
parser.add_argument('--variant', action='store_true', default=False)
parser.add_argument('--seed', type=int, default=42)
parser.add_argument('--learn_rate', type=float, default=1e-2)
parser.add_argument('--weight_decay1', type=float, default=0.01)
parser.add_argument('--weight_decay2', type=float, default=0.0005)
parser.add_argument('--num_epochs', type=int, default=400)
parser.add_argument('--patience', type=int, default=40)
args = parser.parse_args()
graph, features, labels, train_mask, val_mask, test_mask, num_feats, num_classes = load_data_default(
args.dataset)
model = GCNIINet(num_feats,
num_classes,
args.num_hidden,
args.num_layers,
dropout=args.dropout,
alpha=args.alpha,
lamda=args.lamda)
labels = labels.squeeze()
# set_seed(args.seed)
optimizer = th.optim.Adam([{
'params': model.params1,
'weight_decay': args.weight_decay1
}, {
'params': model.params2,
parser.add_argument('--dropout', type=float, default=0.8)
parser.add_argument('--propagation', type=int, default=1)
parser.add_argument('--seed', type=int, default=42)
parser.add_argument('--learn_rate', type=float, default=0.01)
parser.add_argument('--weight_decay', type=float, default=5e-5)
parser.add_argument('--num_epochs', type=int, default=1500)
parser.add_argument('--patience', type=int, default=100)
parser.add_argument('--cuda', type=int, default=0)
parser.add_argument('--filename', type=str, default='VSGC_Multi')
parser.add_argument('--id', type=int, default=0)
args = parser.parse_args()
test_print = False
graph, features, labels, train_mask, val_mask, test_mask,\
num_feats, num_classes = load_data_default(args.dataset)
model = VSGCNetMulti(num_feats,
num_classes,
k=args.k,
dropout=args.dropout,
propagation=args.propagation)
labels = labels.squeeze()
# set_seed(args.seed)
optimizer = th.optim.Adam(model.parameters(),
lr=args.learn_rate,
weight_decay=args.weight_decay)
early_stopping = EarlyStopping(args.patience,
# if t in b:
# count += 1
# print(count)
def remove_reverse_edges(graph):
edges = graph.edges()
m = graph.num_edges()
edges_set = set()
remove_eids = []
for i in range(m):
u, v = edges[0][i].item(), edges[1][i].item()
if u > v:
e = (v, u)
else:
e = (u, v)
if e in edges_set:
remove_eids.append(i)
else:
edges_set.add(e)
graph.remove_edges(th.LongTensor(remove_eids))
graph, features, labels, train_mask, val_mask, test_mask, num_feats, num_classes = load_data_default(
"citeseer")
graph = graph.remove_self_loop()
print(graph.num_edges())
remove_reverse_edges(graph)
print(graph.num_edges())
