-
Notifications
You must be signed in to change notification settings - Fork 0
/
process_isolated_nodes.py
35 lines (31 loc) · 1.3 KB
/
process_isolated_nodes.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
from torch_geometric.utils import contains_isolated_nodes, remove_isolated_nodes
def process_isolated_nodes(edge_index):
if contains_isolated_nodes(edge_index):
new_edge_index, _, mask = remove_isolated_nodes(edge_index)
mapping = {}
for i in range(edge_index.shape[1]):
if edge_index[0, i] != new_edge_index[0, i]:
mapping[new_edge_index[0, i].item()] = edge_index[0, i].item()
return new_edge_index, mapping, mask
else:
return edge_index, None, [True]*(edge_index.shape[1])
def restore_isolated_ndoes(new_edge_index, mapping):
for i in range(new_edge_index.shape[1]):
if new_edge_index[0, i].item() in mapping:
new_edge_index[0, i] = mapping[new_edge_index[0, i].item()]
if new_edge_index[1, i].item() in mapping:
new_edge_index[1, i] = mapping[new_edge_index[1, i].item()]
return new_edge_index
def restore_isolated_ndoes_int(new_edge_index, mask):
mapping={}
i=0
index=0
while index < len(mask):
if mask[index]:
mapping[i]=index
i+=1
index+=1
for i in range(new_edge_index.shape[1]):
new_edge_index[0, i]=mapping[new_edge_index[0, i].item()]
new_edge_index[1, i]=mapping[new_edge_index[1, i].item()]
return new_edge_index