-
Notifications
You must be signed in to change notification settings - Fork 0
/
my_utils.py
180 lines (144 loc) · 7.68 KB
/
my_utils.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
'''
@File :my_utils.py
@Author:Morton
@Date :2020/6/18 16:26
@Desc :some utils which will been used in my project, including: dump and load function,
normalized function and parse_args()
'''
# -*- coding:utf-8 -*-
import gzip
import pickle
import argparse
import numpy as np
import scipy.sparse as ssp
def dump_obj(obj, filename, protocol=-1, serializer=pickle):
with gzip.open(filename, 'wb') as fout:
serializer.dump(obj, fout, protocol)
def load_obj(filename, serializer=pickle):
with gzip.open(filename, 'rb') as fin:
obj = serializer.load(fin)
return obj
def aug_normalized_adjacency(adj):
print("aug_normalized_adjacency of SGC.")
adj = adj + ssp.eye(adj.shape[0])
adj = ssp.coo_matrix(adj)
row_sum = np.array(adj.sum(1))
d_inv_sqrt = np.power(row_sum, -0.5).flatten()
d_inv_sqrt[np.isinf(d_inv_sqrt)] = 0.
d_mat_inv_sqrt = ssp.diags(d_inv_sqrt)
return d_mat_inv_sqrt.dot(adj).dot(d_mat_inv_sqrt).toarray()
def feature_normalization(dt, mode='None'):
'''
normalization according to row, each row represent a feature
'''
print("features normalization: {}".format(mode))
if mode == "Standard":
mean_num = np.mean(dt, axis=0)
sigma = np.std(dt, axis=0)
out = (dt - mean_num) / sigma
if mode == "Mean":
mean_num = np.mean(dt, axis=0)
max_num = np.max(dt, axis=0)
min_num = np.min(dt, axis=0)
out = (dt - mean_num) / (max_num - min_num)
if mode == "None":
out = dt
return out
def load_data_for_SGC(dump_file, feature_norm='None', degree=3):
data = load_obj(dump_file)
adj, X_train, Y_train, X_dev, Y_dev, X_test, Y_test, U_train, U_dev, U_test, \
classLatMedian, classLonMedian, userLocation = data
adj = aug_normalized_adjacency(adj)
features = np.vstack((X_train, X_dev, X_test)) # only for dec2vec_feature
features = feature_normalization(features, mode=feature_norm)
print("features shape: {} , normalization:{}".format(features.shape, feature_norm))
for i in range(0, degree):
features = np.matmul(adj, features)
labels = np.hstack((Y_train, Y_dev, Y_test))
'''get index of train val and test'''
len_train, len_val, len_test = int(X_train.shape[0]), int(X_dev.shape[0]), int(X_test.shape[0])
idx_train, idx_val, idx_test = range(len_train), range(len_train, len_train + len_val), range(len_train + len_val,
len_train + len_val + len_test)
data = (features, labels, idx_train, idx_val, idx_test, U_train, U_dev, U_test, classLatMedian, classLonMedian,
userLocation)
return data
def load_data_for_N2V(dump_file, edge_file, feature_norm='None'):
data = load_obj(dump_file)
adj, X_train, Y_train, X_dev, Y_dev, X_test, Y_test, U_train, U_dev, U_test, \
classLatMedian, classLonMedian, userLocation = data
edge_emb = np.load(edge_file)
print("edge_emb shape: {}".format(edge_emb.shape))
features = np.vstack((X_train, X_dev, X_test)) # only for dec2vec_feature
features = feature_normalization(features, mode=feature_norm)
print("features shape: {} , normalization:{}".format(features.shape, feature_norm))
labels = np.hstack((Y_train, Y_dev, Y_test))
'''get index of train val and test'''
len_train = int(X_train.shape[0])
len_val = int(X_dev.shape[0])
len_test = int(X_test.shape[0])
idx_train = range(len_train)
idx_val = range(len_train, len_train + len_val)
idx_test = range(len_train + len_val, len_train + len_val + len_test)
data = (features, labels, idx_train, idx_val, idx_test, U_train, U_dev,
U_test, classLatMedian, classLonMedian, userLocation, edge_emb)
return data
def load_data_for_plot(dump_file, feature_norm='None'):
data = load_obj(dump_file)
adj, X_train, Y_train, X_dev, Y_dev, X_test, Y_test, U_train, U_dev, U_test, \
classLatMedian, classLonMedian, userLocation = data
features = np.vstack((X_train, X_dev, X_test)) # only for dec2vec_feature
features = feature_normalization(features, mode=feature_norm)
print("features shape: {} , normalization:{}".format(features.shape, feature_norm))
labels = np.hstack((Y_train, Y_dev, Y_test))
'''get index of train val and test'''
len_train = int(X_train.shape[0])
len_val = int(X_dev.shape[0])
len_test = int(X_test.shape[0])
idx_train = range(len_train)
idx_val = range(len_train, len_train + len_val)
idx_test = range(len_train + len_val, len_train + len_val + len_test)
data = (adj, features, labels, idx_train, idx_val, idx_test, U_train, U_dev,
U_test, classLatMedian, classLonMedian, userLocation)
return data
def parse_args(argv):
"""
Parse commandline arguments.
Arguments:
argv -- An argument list without the program name.
"""
parser = argparse.ArgumentParser()
# data_args: control the data loading
parser.add_argument('-dir', metavar='str', help='the detail directory of dataset',
type=str, default='../dataset_cmu/')
parser.add_argument('-dump_file', metavar='str', help='the dir to load file include name',
type=str, default='../dataset_cmu/dump_doc_dim_512.pkl')
parser.add_argument('-feature_norm', type=str, choices=['None', 'Standard', 'Mean'], default='None')
parser.add_argument('-doc2vec_model_file', metavar='str', help='the dir to load doc2vec_model_file .bin',
type=str, default="../dataset_cmu/corpus/model_dim_512_epoch_40.bin")
# For cmu -bucket 50 -celebrity 5
# For na -bucket 2400 -celebrity 15
# For world -bucket 2400 -celebrity 5
# parser.add_argument('-edge_dis_file', metavar='str', help='the dir to load all edges distance file',
# type=str, default='./my_assets/no_ues_now/edge_dis.pkl')
parser.add_argument('-bucket', metavar='int', help='discretisation bucket size', type=int, default=50)
parser.add_argument('-mindf', metavar='int', help='minimum document frequency in BoW', type=int, default=10)
parser.add_argument('-encoding', metavar='str', help='Data Encoding (e.g.latin1, utf-8)', type=str,
default='latin1')
parser.add_argument('-celebrity', metavar='int', help='celebrity threshold', type=int, default=5)
parser.add_argument('-builddata', action='store_true', help='if true do not recreated dumped data', default=False)
# process_args: control the data preprocess
parser.add_argument('-degree', type=int, help='degree of the approximation.', default=4)
parser.add_argument('-normalization', type=str, help='Normalization method for the adjacency matrix.',
choices=['NormLap', 'Lap', 'RWalkLap', 'FirstOrderGCN', 'AugNormAdj', 'NormAdj',
'RWalk', 'AugRWalk', 'NoNorm'], default='AugNormAdj')
# model_args: the hyper-parameter of SGC model
parser.add_argument('-model', type=str, help='model to use.', default="SGC")
parser.add_argument('-usecuda', action='store_true', help='Use CUDA for training.', default=True)
parser.add_argument('-seed', metavar='int', help='random seed', type=int, default=77)
parser.add_argument('-epochs', type=int, help='Number of epochs to train.', default=30000)
parser.add_argument('-lr', type=float, help='Initial learning rate.', default=0.001)
parser.add_argument('-weight_decay', type=float, help='Weight decay (L2 loss on parameters).', default=5e-7)
parser.add_argument('-patience', help='max iter for early stopping', type=int, default=30)
parser.add_argument('-batch', metavar='int', help='SGD batch size', type=int, default=1024)
args = parser.parse_args(argv)
return args