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
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
|
"""Data reading utils."""
import json
import glob
import torch
import numpy as np
import pandas as pd
from tqdm import tqdm
import networkx as nx
from texttable import Texttable
def hierarchical_graph_reader(path):
"""
Reading the macro-level graph from disk.
:param path: Path to the edge list.
:return graph: Hierarchical graph as a NetworkX object.
"""
edges = pd.read_csv(path).values.tolist() # 按行输出
graph = nx.from_edgelist(edges) # 构图
return graph
def graph_level_reader(path):
"""
Reading a single graph from disk.
:param path: Path to the JSON file.
:return data: Dictionary of data.
"""
data = json.load(open(path))
return data
def tab_printer(args):
"""
Function to print the logs in a nice tabular format.
:param args: Parameters used for the model.
"""
args = vars(args) # vars(): 以字典形式返回当前对象的属性和属性值
keys = sorted(args.keys())
t = Texttable()
t.add_rows([["Parameter", "Value"]])
t.add_rows([[k.replace("_", " ").capitalize(), args[k]] for k in keys])
print(t.draw())
class GraphDatasetGenerator(object):
"""
Creating an in memory version of the graphs.
:param path: Folder with json files.
"""
def __init__(self, path, feature_which):
self.path = path
self.feature_which = feature_which
self._enumerate_graphs() # 生成self.graph，label_dict,feature_dict
self._count_features_and_labels() # 计算feature和label的个数
self._create_target() # label的longtensor
self._create_dataset()
def _enumerate_graphs(self):
"""
Listing the graph files and creating the respective label and feature maps.
"""
graph_count = len(glob.glob(self.path + "*.json")) # 查满足条件的文件
labels = set()
features = set()
self.graphs = []
for index in tqdm(range(graph_count)):
graph_file = self._concatenate_name(index) # 拼接文件名称
data = graph_level_reader(graph_file) # load图
for k, v in data["features"].items():
tmp = []
if len(v) > 1:
for i in self.feature_which:
tmp.append(v[i])
data["features"][k] = tmp
self.graphs.append(data)
labels = labels.union(set([int(data["label"])]))
features = features.union(set([val for k, v in data["features"].items() for val in v]))
self.label_map = {v: i for i, v in enumerate(labels)}
self.feature_map = {v: i for i, v in enumerate(features)}
def _count_features_and_labels(self):
"""
Counting the number of unique features and labels.
"""
self.number_of_features = len(self.feature_map)
self.number_of_labels = len(self.label_map)
def _transform_edges(self, raw_data):
"""
Transforming an edge list from the data dictionary to a tensor.
:param raw_data: Dictionary with edge list.
:return : Edge list matrix.
"""
edges = [[edge[0], edge[1]] for edge in raw_data["edge"]]
edges = edges + [[edge[1], edge[0]] for edge in raw_data["edge"]]
return torch.t(torch.LongTensor(edges))
def _concatenate_name(self, index):
"""
Creating a file name from an index.
:param index: Graph index.
:return : File name.
"""
return self.path + str(index) + ".json"
def _transform_features(self, raw_data):
"""
Creating a feature matrix from the raw data.
:param raw_data: Dictionary with features.
:return feature_matrix: FloatTensor of features.
"""
number_of_nodes = len(raw_data["features"])
feature_matrix = np.zeros((number_of_nodes, self.number_of_features))
index_1 = [int(n) for n, feats in raw_data["features"].items() for f in feats] # n节点有几个特征就有几个n [0,0,0,1,1,1]
index_2 = [int(self.feature_map[f]) for n, feats in raw_data["features"].items() for f in feats] # 特征
feature_matrix[index_1, index_2] = 1.0
feature_matrix = torch.FloatTensor(feature_matrix)
return feature_matrix
def _data_transform(self, raw_data):
"""
Creating a dictionary with the edge list matrix and the features matrix.
"""
clean_data = dict()
clean_data["edge"] = self._transform_edges(raw_data)
clean_data["features"] = self._transform_features(raw_data)
return clean_data
def _create_target(self):
"""
Creating a target vector.
"""
self.type_ind = [[] for i in range(len(self.label_map))]
self.target = []
for i in range(len(self.graphs)):
self.type_ind[int(self.graphs[i]["label"])].append(i)
self.target.append(int(self.graphs[i]["label"]))
# self.target = [int(graph["label"]) for graph in self.graphs]
self.target = torch.LongTensor(self.target)
def _create_dataset(self):
"""
Creating a list of dictionaries with edge list matrices and feature matrices.
"""
self.graphs = [self._data_transform(graph) for graph in self.graphs] # dict edges： features：
|
