forked from layer6ai-labs/DropoutNet
-
Notifications
You must be signed in to change notification settings - Fork 0
/
data.py
149 lines (127 loc) · 6.56 KB
/
data.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
import numpy as np
import torch
import scipy.sparse
import utils
import pandas as pd
"""
This module contains class and methods related to data used in DropoutNet
"""
def load_eval_data(test_file, test_id_file, name, cold, train_data, citeu=False):
timer = utils.timer(name='utils')
with open(test_id_file) as f:
test_item_ids = [int(line) for line in f]
test_data = pd.read_csv(test_file, delimiter=",", header=None, dtype=np.int32).values.ravel()
if citeu:
test_data = test_data.view(
dtype=[('uid', np.int32), ('iid', np.int32), ('inter', np.int32)])
else:
test_data = test_data.view(
dtype=[('uid', np.int32), ('iid', np.int32), ('inter', np.int32), ('date', np.int32)])
timer.toc('read %s triplets %s' % (name, test_data.shape)).tic()
eval_data = EvalData(
test_data,
test_item_ids,
is_cold=cold,
train=train_data
)
timer.toc('loaded %s' % name).tic()
print(eval_data.get_stats_string())
return eval_data
class EvalData:
"""
EvalData:
EvalData packages test triplet (user, item, score) into appropriate formats for evaluation
Compact Indices:
Specifically, this builds compact indices and stores mapping between original and compact indices.
Compact indices only contains:
1) items in test set
2) users who interacted with such test items
These compact indices speed up testing significantly by ignoring irrelevant users or items
Args:
test_triplets(int triplets): user-item-interaction_value triplet to build the test data
train(int triplets): user-item-interaction_value triplet from train data
Attributes:
is_cold(boolean): whether test data is used for cold start problem
test_item_ids(list of int): maps compressed item ids to original item ids (via position)
test_item_ids_map(dictionary of int->int): maps original item ids to compressed item ids
test_user_ids(list of int): maps compressed user ids to original user ids (via position)
test_user_ids_map(dictionary of int->int): maps original user ids to compressed user ids
R_test_inf(scipy lil matrix): pre-built compressed test matrix
R_train_inf(scipy lil matrix): pre-built compressed train matrix for testing
other relevant input/output exposed from tensorflow graph
"""
def __init__(self, test_triplets, test_item_ids, is_cold, train):
# build map both-ways between compact and original indices
# compact indices only contains:
# 1) items in test set
# 2) users who interacted with such test items
self.is_cold = is_cold
self.test_item_ids = test_item_ids
# test_item_ids_map
self.test_item_ids_map = {iid: i for i, iid in enumerate(self.test_item_ids)}
_test_ij_for_inf = [(t[0], t[1]) for t in test_triplets if t[1] in self.test_item_ids_map]
# test_user_ids
self.test_user_ids = np.unique(test_triplets['uid'])
# test_user_ids_map
self.test_user_ids_map = {user_id: i for i, user_id in enumerate(self.test_user_ids)}
_test_i_for_inf = [self.test_user_ids_map[_t[0]] for _t in _test_ij_for_inf]
_test_j_for_inf = [self.test_item_ids_map[_t[1]] for _t in _test_ij_for_inf]
self.R_test_inf = scipy.sparse.coo_matrix(
(np.ones(len(_test_i_for_inf)),
(_test_i_for_inf, _test_j_for_inf)),
shape=[len(self.test_user_ids), len(self.test_item_ids)]
).tolil(copy=False)
train_ij_for_inf = [(self.test_user_ids_map[_t[0]], self.test_item_ids_map[_t[1]]) for _t
in train
if _t[1] in self.test_item_ids_map and _t[0] in self.test_user_ids_map]
if self.is_cold and len(train_ij_for_inf) != 0:
raise Exception('using cold dataset, but data is not cold!')
if not self.is_cold and len(train_ij_for_inf) == 0:
raise Exception('using warm datset, but data is not warm!')
self.R_train_inf = None if self.is_cold else scipy.sparse.coo_matrix((
np.ones(len(train_ij_for_inf)),
zip(*train_ij_for_inf)), shape=self.R_test_inf.shape).tolil(copy=False)
# allocate fields
self.U_pref_test = None
self.V_pref_test = None
self.V_content_test = None
self.U_content_test = None
self.tf_eval_train = None
self.tf_eval_test = None
self.eval_batch = None
def init_tf(self, user_factors, item_factors, user_content, item_content, eval_run_batchsize):
self.U_pref_test = user_factors[self.test_user_ids, :]
self.V_pref_test = item_factors[self.test_item_ids, :]
self.V_content_test = item_content[self.test_item_ids, :]
if scipy.sparse.issparse(self.V_content_test):
self.V_content_test = self.V_content_test.todense()
if user_content!=None:
self.U_content_test = user_content[self.test_user_ids, :]
if scipy.sparse.issparse(self.U_content_test):
self.U_content_test = self.U_content_test.todense()
eval_l = self.R_test_inf.shape[0]
self.eval_batch = [(x, min(x + eval_run_batchsize, eval_l)) for x
in range(0, eval_l, eval_run_batchsize)]
self.tf_eval_train = []
self.tf_eval_test = []
if not self.is_cold:
for (eval_start, eval_finish) in self.eval_batch:
_ui = self.R_train_inf[eval_start:eval_finish, :].tocoo()
_ui = list(zip(_ui.row, _ui.col))
self.tf_eval_train.append(
torch.sparse_coo_tensor(
indices=np.array(_ui).T,
values=np.full(len(_ui), -100000, dtype=np.float32),
size=[eval_finish - eval_start, self.R_train_inf.shape[1]]
)
)
def get_stats_string(self):
return ('\tn_test_users:[%d]\n\tn_test_items:[%d]' % (len(self.test_user_ids), len(self.test_item_ids))
+ '\n\tR_train_inf: %s' % (
'no R_train_inf for cold' if self.is_cold else 'shape=%s nnz=[%d]' % (
str(self.R_train_inf.shape), len(self.R_train_inf.nonzero()[0])
)
)
+ '\n\tR_test_inf: shape=%s nnz=[%d]' % (
str(self.R_test_inf.shape), len(self.R_test_inf.nonzero()[0])
))