-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtraining_pipeline.py
178 lines (151 loc) · 7.37 KB
/
training_pipeline.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
from utils import *
from loss import *
import matplotlib.cm as cm
import matplotlib.pyplot as plt
from config import InitLearningRate,Args
from train import startTraining
def EntryPoint(mode='training',cuda=1,gpu=0):
test_number = 10
work_dir = os.getcwd()
nn_model = 'GCN'
data_dir = os.path.join(work_dir, 'data/ComboSampleData/')
##variables to store final result
modularity_scores_gcn = {}
nmi_gcn={}
nmi={}
C_init={}
C_out = {}
C_out_combo = {}
graph_type = {}
n_communities={}
initial_partition_approach={}
modularity_scores_combo = {}
modularity_scores_combo_restricted={}
loss={}
model_parameter = {}
data_name = []
modularity_scores_classic={}
training_loops = 100
result_score = []
for i in range(training_loops):
for root, dirs, files in os.walk(data_dir):
for file in files:
if file[-3:] == 'mat':
#append dataset name list
if 'karate_34' not in file:
continue
dataset = file[:-4]
data_name.append(dataset )
G = loadNetworkMat(file, data_dir)
if nx.classes.function.is_directed(G):
graph_type[file] = 'directed'
initial_partition_approach[file]='LPA'
else:
graph_type[file] = 'undirected'
initial_partition_approach[file]='Louvain'
print(file, graph_type[file])
if mode == 'training':
lr = InitLearningRate(dataset,use_Adam=False,use_default_lr=False)
# construct args as training parameter
args = Args(dataset=dataset)
args.setArgs(cuda=cuda,
grad_direction=-1,
nn_model=nn_model)
args.setArgs(
learning_rate=lr.get_init_lr(),
lr_mode=mode,
n_epochs=15000,
step_size=100,
early_stop=True
)
elif mode=='scanning':
lr = InitLearningRate(dataset)
# construct args as training parameter
args = Args(dataset=dataset)
args.setArgs(cuda=cuda,
grad_direction=-1,
nn_model=nn_model)
args.setArgs(
learning_rate=lr.get_init_lr(),
lr_mode=mode,
n_epochs=150,
step_size=1,
)
modularity_scores_combo[file], partition = getNewComboPartition(G)
C_out_combo[file] = partition_to_binary_attachment(partition)
modularity_scores_gcn[file], loss[file], C_out[file], model_parameter[file], C_init[file], \
modularity_scores_classic[file], n_communities[file] = startTraining(nx_g=G,data_dir=data_dir,dataset=dataset,args=args.getArgs())
modularity_scores_combo_restricted[file], partition = getNewComboPartition(G, maxcom=n_communities[file])
nmi_gcn[file]=NMI(C_out[file],C_init[file])
nmi[file] = NMI(C_out_combo[file], C_init[file])
result_score.append( loss[file])
##save log
# save_result(data_name, graph_type, modularity_scores_gcn, modularity_scores_combo, modularity_scores_combo_restricted,modularity_scores_classic,nmi_gcn,nmi,model_parameter,
# data_dir)
# #<network>,<initial partition approach>,<number of communities>,<initial modularity>,<modularity after fune-tuning>,<COMBO modularity >,<COMBO modularity without restricting the number of communities and the optimal number of communities it returns >
# save_result_for_report(data_name, initial_partition_approach,n_communities, modularity_scores_gcn, loss,modularity_scores_combo, modularity_scores_combo_restricted,modularity_scores_classic,nmi_gcn,nmi,model_parameter,
# data_dir)
print(sum(result_score)/len(result_score))
print('something')
def temp():
'''
test_number = 10
work_dir = os.getcwd()
nn_model = 'GCN'
data_dir = os.path.join(work_dir, 'data/ComboSampleData/')
# G = loadNetworkMat('karate_34.mat',data_dir)
G = loadNetworkMat('celeganmetabolic_453.mat', data_dir)
modularity_scores_gcn = {}
nmi_gcn={}
nmi={}
C_init={}
C_out = {}
C_out_combo = {}
graph_type = {}
n_communities={}
initial_partition_approach={}
modularity_scores_combo = {}
modularity_scores_combo_restricted={}
loss={}
model_parameter = {}
data_name = []
modularity_scores_classic={}
for root, dirs, files in os.walk(data_dir):
for file in files:
# print(file[-3:])
# if 'celegansneural_297' not in file:
# continue
if file[-3:] == 'mat':
#append dataset name list
# if 'karate_34' not in file:
# continue
data_name.append(file)
G = loadNetworkMat(file, data_dir)
if nx.classes.function.is_directed(G):
graph_type[file] = 'directed'
initial_partition_approach[file]='LPA'
else:
graph_type[file] = 'undirected'
initial_partition_approach[file]='Louvain'
print(file, graph_type[file])
# need to figure out it is weighted or not
modularity_scores_combo[file], partition = getNewComboPartition(G)
C_out_combo[file] = partition_to_binary_attachment(partition)
#modularity_score, C_hat, model_structure, features, n_classes
modularity_scores_gcn[file], loss[file],C_out[file], model_parameter[file],C_init[file],modularity_scores_classic[file], n_communities[file] = main(G, nn_model,grad_direction=-1,data_dir=data_dir,dataset=file[:-4],args)
modularity_scores_combo_restricted[file], partition = getNewComboPartition(G, maxcom=n_communities[file])
nmi_gcn[file]=NMI(C_out[file],C_init[file])
nmi[file] = NMI(C_out_combo[file], C_init[file])
##save log
save_result(data_name, graph_type, modularity_scores_gcn, modularity_scores_combo, modularity_scores_combo_restricted,modularity_scores_classic,nmi_gcn,nmi,model_parameter,
data_dir)
#<network>,<initial partition approach>,<number of communities>,<initial modularity>,<modularity after fune-tuning>,<COMBO modularity >,<COMBO modularity without restricting the number of communities and the optimal number of communities it returns >
save_result_for_report(data_name, initial_partition_approach,n_communities, modularity_scores_gcn, loss,modularity_scores_combo, modularity_scores_combo_restricted,modularity_scores_classic,nmi_gcn,nmi,model_parameter,
data_dir)
print('something')
:return:
'''
if __name__ == "__main__":
cuda=1
gpu=0
EntryPoint(mode='training',cuda=cuda,gpu=gpu)