-
Notifications
You must be signed in to change notification settings - Fork 1
/
utility.py
194 lines (147 loc) · 7.01 KB
/
utility.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
#coding: utf-8
import os
import math
import json
import random
random.seed(1234)
import numpy as np
from datetime import datetime
import scipy.sparse as sp
import torch
from torch.utils.data import Dataset, DataLoader
def sample_index(ttl_num):
f1 = int(ttl_num / 10000)
f2 = int((ttl_num - f1 * 10000) / 100)
f3 = int((ttl_num - f1 * 10000 - f2 * 100))
ind1 = random.randint(0, f1)
if ind1 == f1:
ind23 = random.randint(0, f2)
if ind23 == f2:
ind45 = random.randint(0, f3 - 1)
else:
ind45 = random.randint(0, 99)
else:
ind23 = random.randint(0, 99)
ind45 = random.randint(0, 99)
ind = ind1 * 10000 + ind23 * 100 + ind45
return ind
class TrainData(Dataset):
def __init__(self, conf, input_seq, user_seq, item_seq, all_items):
self.conf = conf
self.input_seq = input_seq
self.user_seq = user_seq
self.item_seq = item_seq
self.n_all = len(all_items)
def __len__(self):
return len(self.input_seq)
def __getitem__(self, idx):
seq = self.input_seq[idx]
u = seq[-1]
i = seq[-2]
l = seq[-3]
hist = seq[:-2]
user_seq = self.user_seq[str(u)]
item_seq = self.item_seq[str(l)]
k = sample_index(self.n_all)
while k in user_seq or k in item_seq:
k = sample_index(self.n_all)
return u, l, i, int(k)
class TestData(Dataset):
def __init__(self, conf, input_seq, neg_seq):
self.conf = conf
self.input_seq = input_seq
self.neg_seqs = neg_seq # list of neg samples for each item
def __len__(self):
return len(self.input_seq)
def __getitem__(self, idx):
seq = self.input_seq[idx]
u = seq[-1]
i = seq[-2]
l = seq[-3]
hist = seq[:-2]
ks = [self.neg_seqs[cnt][idx] for cnt in self.neg_seqs.keys()]
ks = torch.LongTensor(ks)
return u, l, i, ks
class DGSR_Dataset():
def __init__(self, conf):
all_data = self.load_cache_data(conf)
self.train_seqs, self.val_seqs, self.val_negs, self.test_seqs, self.test_negs, self.user_id_map, self.id_user_map, self.item_id_map, self.id_item_map, self.user_item_set, self.item_item_set = all_data
self.item_ids = list(self.id_item_map.keys())
self.train_pairs_ui = self.load_cf_data()
self.train_pairs_ii = self.load_trans_data()
self.n_users = len(self.user_id_map)
self.n_items = len(self.item_id_map)
self.train_len = len(self.train_seqs)
self.adj_ui, self.adj_iu = self.get_adj(self.train_pairs_ui, self.n_users, self.n_items)
self.adj_ij, self.adj_ji = self.get_adj(self.train_pairs_ii, self.n_items, self.n_items)
self.train_set = TrainData(conf, self.train_seqs, self.user_item_set, self.item_item_set, self.item_ids)
self.train_loader = DataLoader(self.train_set, batch_size=conf["batch_size"], shuffle=True, num_workers=10)
self.test_set = TestData(conf, self.test_seqs, self.test_negs)
self.test_loader = DataLoader(self.test_set, batch_size=conf["test_batch_size"], shuffle=False, num_workers=10)
self.val_set = TestData(conf, self.val_seqs, self.val_negs)
self.val_loader = DataLoader(self.val_set, batch_size=conf["test_batch_size"], shuffle=False, num_workers=10)
def load_cf_data(self):
train_pairs = []
for train_seq in self.train_seqs:
uid = train_seq[-1]
iid = train_seq[-2]
train_pairs.append([uid, iid])
return np.array(train_pairs)
def load_trans_data(self):
train_pairs = []
for train_seq in self.train_seqs:
iid = train_seq[-2]
lid = train_seq[-3]
train_pairs.append([lid, iid])
return np.array(train_pairs)
def get_adj(self, train_pairs, n_node1, n_node2):
a_rows = train_pairs[:, 0]
a_cols = train_pairs[:, 1]
a_vals = [1.] * len(a_rows)
b_rows = a_cols
b_cols = a_rows
b_vals = [1.] * len(b_rows)
a_adj = sp.coo_matrix((a_vals, (a_rows, a_cols)), shape=(n_node1, n_node2))
b_adj = sp.coo_matrix((b_vals, (b_rows, b_cols)), shape=(n_node2, n_node1))
a_adj = self.get_lap(a_adj)
b_adj = self.get_lap(b_adj)
return a_adj, b_adj
def get_lap(self, adj):
def bi_norm_lap(adj):
rowsum = np.array(adj.sum(1))
d_inv_sqrt = np.power(rowsum + 0.00000001, -0.5).flatten()
d_inv_sqrt[np.isinf(d_inv_sqrt)] = 0.
d_mat_inv_sqrt = sp.diags(d_inv_sqrt)
bi_lap = d_mat_inv_sqrt.dot(adj).dot(d_mat_inv_sqrt)
return bi_lap.tocoo()
def si_norm_lap(adj):
rowsum = np.array(adj.sum(1))
d_inv = np.power(rowsum + 0.00000001, -1).flatten()
d_inv[np.isinf(d_inv)] = 0.
d_mat_inv = sp.diags(d_inv)
norm_adj = d_mat_inv.dot(adj)
return norm_adj.tocoo()
lap = si_norm_lap(adj)
return lap
def load_cache_data(self, conf):
target_path = conf["target_path"]
val_negs, test_negs = {}, {}
train_seqs = json.load(open(target_path + "train_seqs_%d_large.json"%conf['seq_len']))
if conf["dataset"] in ["ifashion"]: # val_negs and test_negs for ifashion is a dict with more than one set of negs, we take the first set.
[val_seqs, val_negs[0]] = json.load(open(target_path + "val_seqs_%d.json"%conf['seq_len']))
[test_seqs, test_negs[0]] = json.load(open(target_path + "test_seqs_%d.json"%conf['seq_len']))
else: # val_negs and test_negs for amazon is a list, which is the only neg set
val_seqs = json.load(open(target_path + "val_seqs_%d.json"%conf['seq_len']))
val_negs = json.load(open(target_path + "val_negs_%d.json"%conf['seq_len']))
test_seqs= json.load(open(target_path + "test_seqs_%d.json"%conf['seq_len']))
test_negs = json.load(open(target_path + "test_negs_%d.json"%conf['seq_len']))
user_id_map = json.load(open(target_path + "user_id_map_%d.json"%conf['seq_len']))
id_user_map = json.load(open(target_path + "id_user_map_%d.json"%conf['seq_len']))
item_id_map = json.load(open(target_path + "item_id_map_%d.json"%conf['seq_len']))
id_item_map = json.load(open(target_path + "id_item_map_%d.json"%conf['seq_len']))
user_item_set = json.load(open(target_path + "user_item_set_%d.json"%conf['seq_len']))
item_item_set = json.load(open(target_path + "item_pos_item_set_%d.json"%conf['seq_len']))
# convert values into sets to accelerrate the lookup
user_item_set = {user: set(items) for user, items in user_item_set.items()}
item_item_set = {item: set(items) for item, items in item_item_set.items()}
return train_seqs, val_seqs, val_negs, test_seqs, test_negs, user_id_map, id_user_map, item_id_map, id_item_map, user_item_set, item_item_set