/
pipline.py
197 lines (191 loc) · 9.53 KB
/
pipline.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
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
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
import argparse
import warnings
import os
import torch
import numpy as np
from pytorch_HNECV import getdataloader,train_VAE
from RandomWalk import GeneratorMetaPath_by_randomwalk
from RandomWalk_Yelp import GeneratorMetaPath_by_randomwalk_yelp
from utils import Dataset,build_fusion_mat,select_embeddingBylabel
from Node_classifier import preprocess_for_knn
from Node_clustering import preprocess_for_kmeans
from TSNE import preprocess_for_tsne
def start(args):
dataset = Dataset()
if args.dataset == 'dblp':
dataset.preprocess_dblp()
init_class = GeneratorMetaPath_by_randomwalk()
# Initial data set path
dirpath = "dataset/DBLP/reindex_dblp/"
# output path
output_path = "dataset/DBLP/dblp_metapath_by_rw/"
# Number of rows generated by a single start node
numwalks = 10
# walk length
walklength = 80
# Save the meta path sequence file path
outputfile_APA = output_path + "APA_metapath_DBLP_r" + str(
numwalks) + "_c" + str(walklength) + ".txt"
outputfile_APCPA = output_path + "APCPA_metapath_DBLP_r" + str(
numwalks) + "_c" + str(walklength) + ".txt"
outputfile_APTPA = output_path + "APTPA_metapath_DBLP_r" + str(
numwalks) + "_c" + str(walklength) + ".txt"
init_class.build_adj_dict(dirpath, 'paper_author_new.txt', init_class.paper_author, init_class.author_paper)
init_class.build_adj_dict(dirpath, 'paper_conf_new.txt', init_class.paper_conf, init_class.conf_paper)
init_class.build_adj_dict(dirpath, 'paper_term_new.txt', init_class.paper_term, init_class.term_paper)
print("start random walk.....")
init_class.parallel_random_walk(init_class.random_walk_by_APA, outputfile_APA, numwalks, walklength)
init_class.parallel_random_walk(init_class.random_walk_by_APCPA, outputfile_APCPA, numwalks,
walklength)
init_class.parallel_random_walk(init_class.random_walk_by_APTPA, outputfile_APTPA, numwalks,
walklength)
dataset = Dataset()
dataset.parallel_process(output_path)
graph_mat_path = 'dataset/DBLP/dblp_graph_mat/'
dataset.build_graph_from_rwfile_mat(graph_mat_path, output_path)
build_fusion_mat(graph_mat_path,args.dataset)
if args.dataset == 'aminer':
dataset.preprocess_aminer()
init_class = GeneratorMetaPath_by_randomwalk()
# Initial data set path
dirpath = "dataset/AMiner/reindex_aminer/"
# output path
output_path = "dataset/AMiner/aminer_metapath_by_rw/"
# Number of rows generated by a single start node
numwalks = 10
# walk length
walklength = 80
# Save the meta path sequence file path
outputfile_APA = output_path + "APA_metapath_Aminer_r" + str(
numwalks) + "_c" + str(walklength) + ".txt"
outputfile_APCPA = output_path + "APCPA_metapath_Aminer_r" + str(
numwalks) + "_c" + str(walklength) + ".txt"
init_class.build_adj_dict(dirpath, 'paper_author_new.txt', init_class.paper_author, init_class.author_paper)
init_class.build_adj_dict(dirpath, 'paper_conf_new.txt', init_class.paper_conf, init_class.conf_paper)
print("start random walk.....")
init_class.parallel_random_walk(init_class.random_walk_by_APA, outputfile_APA, numwalks, walklength)
init_class.parallel_random_walk(init_class.random_walk_by_APCPA, outputfile_APCPA, numwalks,
walklength)
dataset = Dataset()
dataset.parallel_process(output_path)
graph_mat_path = 'dataset/AMiner/aminer_graph_mat/'
dataset.build_graph_from_rwfile_mat(graph_mat_path, output_path)
build_fusion_mat(graph_mat_path, args.dataset)
if args.dataset == 'yelp':
dataset.preprocess_yelp()
init_class = GeneratorMetaPath_by_randomwalk_yelp()
# Initial data set path
dirpath = "dataset/Yelp/sourcedata/"
# output path
output_path = "dataset/Yelp/yelp_metapath_by_rw/"
# Number of rows generated by a single start node
numwalks = 10
# walk length
walklength = 80
# Save the meta path sequence file path
outputfile_BSB = output_path+"BSB_metapath_Yelp_r" + str(numwalks) + "_c" + str(
walklength) + ".txt"
outputfile_BStB = output_path+"BStB_metapath_Yelp_r" + str(numwalks) + "_c" + str(
walklength) + ".txt"
outputfile_BUB = output_path+"BUB_metapath_Yelp_r" + str(numwalks) + "_c" + str(
walklength) + ".txt"
init_class.build_adj_dict(dirpath, 'business_service.txt', init_class.business_service,
init_class.service_business)
init_class.build_adj_dict(dirpath, 'business_stars.txt', init_class.business_stars, init_class.stars_business)
init_class.build_adj_dict(dirpath, 'business_user.txt', init_class.business_user, init_class.user_business)
print("start random walk.....")
init_class.parallel_random_walk(init_class.random_walk_by_BSB, outputfile_BSB, numwalks, walklength)
init_class.parallel_random_walk(init_class.random_walk_by_BStB, outputfile_BStB, numwalks, walklength)
init_class.parallel_random_walk(init_class.random_walk_by_BUB, outputfile_BUB, numwalks, walklength)
dataset = Dataset()
dataset.parallel_process(output_path)
graph_mat_path = 'dataset/Yelp/yelp_graph_mat/'
dataset.build_graph_from_rwfile_mat(graph_mat_path, output_path)
build_fusion_mat(graph_mat_path, args.dataset)
N, data_loader, data_loader2 = getdataloader(args.input_path, args.batch,device)
print(args)
z = train_VAE(N, args.lr, args.n_epochs, args.pretrain_epochs, args.batch, args.M,
args.loss_path, args.model_path,
args.pretrain_path, args.beta, data_loader, data_loader2,device)
if args.isValidate == True:
selectedEmbedding, selectedlabel = select_embeddingBylabel(z, args.label_path)
if args.savembedding == True:
np.savetxt(args.embedding_path, selectedEmbedding, delimiter=',', fmt='%.07f')
preprocess_for_knn(selectedEmbedding, selectedlabel)
preprocess_for_kmeans(selectedEmbedding, selectedlabel, args.M)
if args.dataset == 'dblp':
preprocess_for_tsne(selectedEmbedding, selectedlabel)
else:
np.savetxt(args.embedding_path, z, delimiter=',', fmt='%.07f')
if __name__ == '__main__':
warnings.filterwarnings("ignore")
os.environ["CUDA_VISIBLE_DEVICES"] = "1"
parser = argparse.ArgumentParser(
description='pipline',
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('--dataset', type=str, default='dblp')
parser.add_argument('--n_epochs', type=int, default=100)
parser.add_argument('--pretrain_epochs', type=int, default=30)
parser.add_argument('--batch', type=int, default=32)
parser.add_argument('--M', type=int, default=4)
parser.add_argument('--beta', type=int, default=2)
parser.add_argument('--lr', type=float, default=1e-4)
parser.add_argument('--dim', type=int, default=128)
parser.add_argument('--pretrain_path', type=str)
parser.add_argument('--input_path', type=str)
parser.add_argument('--embedding_path', type=str)
parser.add_argument('--loss_path', type=str)
parser.add_argument('--model_path', type=str)
parser.add_argument('--label_path', type=str)
parser.add_argument('--isValidate', type=bool, default=True)
parser.add_argument('--savembedding', type=bool, default=True)
args = parser.parse_args()
args.cuda = torch.cuda.is_available()
print("use cuda: {}".format(args.cuda))
device = torch.device("cuda" if args.cuda else "cpu")
if args.dataset == 'dblp':
args.M = 4
args.lr = 0.0001
args.batch = 128
args.beta = 2
args.n_epochs = 100
args.pretrain_epochs = 30
args.pretrain_path = "./pretrain_ae.pkl"
args.input_path = "dataset/DBLP/test/Single_DBLP_mat.mat"
args.embedding_path = "modelResult/DBLP/dblp_embedding.txt"
args.loss_path = "loss.txt"
args.model_path = "modelResult/DBLP/dblp.pkl"
args.label_path = "dataset/DBLP/reindex_dblp/author_label_new.txt"
args.isValidate = True
args.savembedding = True
if args.dataset == 'aminer':
args.M = 8
args.lr = 0.0001
args.batch = 128
args.beta = 5
args.n_epochs = 100
args.pretrain_epochs = 30
args.pretrain_path = "./pretrain_ae.pkl"
args.input_path = "dataset/AMiner/test/Single_Aminer_mat.mat"
args.embedding_path = "modelResult/AMiner/aminer_embedding.txt"
args.loss_path = "loss.txt"
args.model_path = "modelResult/AMiner/aminer.pkl"
args.label_path = "dataset/AMiner/reindex_aminer/author_label_new.txt"
args.isValidate = True
args.savembedding = True
if args.dataset == 'yelp':
args.M = 3
args.lr = 0.0003
args.batch = 32
args.beta = 15
args.n_epochs = 100
args.pretrain_epochs = 50
args.pretrain_path = "./pretrain_ae.pkl"
args.input_path = "dataset/Yelp/test/Single_Yelp_mat.mat"
args.embedding_path = "modelResult/Yelp/yelp_embedding.txt"
args.loss_path = "loss.txt"
args.model_path = "modelResult/Yelp/yelp.pkl"
args.label_path = "dataset/Yelp/entity/business_label.txt"
args.isValidate = True
args.savembedding = True
start(args)