forked from RUCAIBox/RecBole-CDR
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdcdcsr.py
280 lines (251 loc) · 14.2 KB
/
dcdcsr.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
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
# -*- coding: utf-8 -*-
# @Time : 2022/5/21
# @Author : Shanlei Mu
# @Email : [email protected]
r"""
DCDCSR
################################################
Reference:
Feng Zhu et al. "A Deep Framework for Cross-Domain and Cross-System Recommendations" in IJCAI 2018.
"""
import torch
import torch.nn as nn
import numpy as np
from recbole.utils import InputType
from recbole.model.init import xavier_normal_initialization
from recbole.model.layers import MLPLayers
from recbole.model.loss import BPRLoss
from recbole_cdr.model.crossdomain_recommender import CrossDomainRecommender
class DCDCSR(CrossDomainRecommender):
r"""DCDCSR utilizes the sparsity degrees of individual users and items
in the source and target domains to learn an mapping function from source
latent space to target latent space.
"""
input_type = InputType.PAIRWISE
def __init__(self, config, dataset):
super(DCDCSR, self).__init__(config, dataset)
assert self.overlapped_num_items == 1 or self.overlapped_num_users == 1, \
"DCDCSR model only support user overlapped or item overlapped dataset! "
if self.overlapped_num_users > 1:
self.mode = 'overlap_users'
elif self.overlapped_num_items > 1:
self.mode = 'overlap_items'
else:
self.mode = 'non_overlap'
self.phase = None
self.phase2count = {'SOURCE': 0, 'TARGET': 0, 'BOTH': 0, 'OVERLAP': 0}
# load parameters info
self.latent_factor_model = config['latent_factor_model']
assert self.latent_factor_model in ['BPR'],\
"latent_factor model must be in [BPR]"
self.embedding_size = config['embedding_size']
self.mlp_hidden_size = config['mlp_hidden_size']
self.k = config['k']
self.map_batch_size = config['map_batch_size']
# load dataset info
self.SOURCE_LABEL = dataset.source_domain_dataset.label_field
self.TARGET_LABEL = dataset.target_domain_dataset.label_field
if self.mode == 'overlap_items':
self.source_item2pop = self.build_unit2pop(dataset, unit='item', domain='source')
self.target_item2pop = self.build_unit2pop(dataset, unit='item', domain='target')
elif self.mode == 'overlap_users':
self.source_user2pop = self.build_unit2pop(dataset, unit='user', domain='source')
self.target_user2pop = self.build_unit2pop(dataset, unit='user', domain='target')
# define layers and loss
self.source_user_embedding = nn.Embedding(self.total_num_users, self.embedding_size)
self.source_item_embedding = nn.Embedding(self.total_num_items, self.embedding_size)
self.target_user_embedding = nn.Embedding(self.total_num_users, self.embedding_size)
self.target_item_embedding = nn.Embedding(self.total_num_items, self.embedding_size)
self.benchmark_embedding = None
self.affine_embedding = None
with torch.no_grad():
self.source_user_embedding.weight[self.overlapped_num_users: self.target_num_users].fill_(0)
self.source_item_embedding.weight[self.overlapped_num_items: self.target_num_items].fill_(0)
self.target_user_embedding.weight[self.target_num_users:].fill_(0)
self.target_item_embedding.weight[self.target_num_items:].fill_(0)
self.mapping_mlp_layers = MLPLayers(layers=[self.embedding_size] + self.mlp_hidden_size + [self.embedding_size],
activation='tanh', dropout=0, bn=False)
self.rec_loss = None
if self.latent_factor_model == 'BPR':
self.rec_loss = BPRLoss()
self.map_loss = nn.MSELoss()
# parameters initialization
self.apply(xavier_normal_initialization)
@staticmethod
def build_unit2pop(dataset, unit='user', domain='source'):
if unit == 'user':
_, _, history_lens = dataset.history_item_matrix(domain=domain)
else:
_, _, history_lens = dataset.history_user_matrix(domain=domain)
return history_lens.float()
def set_phase(self, phase):
self.phase = phase
self.phase2count[phase] += 1
if self.phase == 'BOTH':
self.build_benchmark_embedding()
if self.phase == 'TARGET' and self.phase2count[self.phase] == 2:
if self.mode == 'overlap_users':
target_user_embedding, mean_, max_ = \
self.maxmin_normalize(self.target_user_embedding.weight[:self.target_num_users])
self.affine_embedding = self.mapping_mlp_layers(
target_user_embedding)
self.affine_embedding = self.affine_embedding * (max_ - mean_) + mean_
self.affine_embedding = self.affine_embedding.detach()
elif self.mode == 'overlap_items':
target_item_embedding, mean_, max_ = \
self.maxmin_normalize(self.target_item_embedding.weight[:self.target_num_items])
self.affine_embedding = self.mapping_mlp_layers(
target_item_embedding)
self.affine_embedding = self.affine_embedding * (max_ - mean_) + mean_
self.affine_embedding = self.affine_embedding.detach()
def calculate_rec_loss(self, interaction, user_embeds, item_embeds,
user_field, item_field, neg_item_field, label_field):
loss = None
if self.latent_factor_model == 'BPR':
user = interaction[user_field]
item = interaction[item_field]
neg_item = interaction[neg_item_field]
user_e = user_embeds[user]
item_e = item_embeds[item]
neg_item_e = item_embeds[neg_item]
pos_score = torch.mul(user_e, item_e).sum(dim=1)
neg_score = torch.mul(user_e, neg_item_e).sum(dim=1)
loss = self.rec_loss(pos_score, neg_score)
return loss
def build_unit_benchmark_embedding(self, total_num_units, overlapped_num_units,
source_unit2pop, target_unit2pop,
source_unit_embeddings, target_unit_embeddings):
self.benchmark_embedding = torch.FloatTensor(total_num_units, self.embedding_size).to(self.device)
for idx in range(overlapped_num_units):
denominator = source_unit2pop[idx] + target_unit2pop[idx]
if denominator == 0:
denominator = 1
alpha_s = source_unit2pop[idx] / denominator
alpha_t = 1 - alpha_s
self.benchmark_embedding[idx] = (alpha_s * target_unit_embeddings[idx] +
alpha_t * source_unit_embeddings[idx]).detach()
for idx in range(overlapped_num_units, total_num_units):
i_e = target_unit_embeddings[idx] # (embedding_size)
sim_i = torch.mm(source_unit_embeddings, i_e.unsqueeze(1)).squeeze(1) # (n_items)
sim, index = torch.topk(sim_i, k=self.k, dim=0) # (k)
sn = torch.mean(source_unit2pop[index])
beta_t_i = sn / (sn + target_unit2pop[idx])
sim_e = source_unit_embeddings[index] # (k, embedding_size)
sim_e = torch.mm(sim.unsqueeze(0), sim_e).squeeze(0) # (embedding_size)
sum_sim = torch.sum(sim) if torch.sum(sim) > 0 else 1
sim_e = sim_e / sum_sim
self.benchmark_embedding[idx] = ((1 - beta_t_i) * target_unit_embeddings[idx] +
beta_t_i * sim_e).detach()
def build_benchmark_embedding(self):
if self.mode == 'overlap_users':
self.build_unit_benchmark_embedding(self.total_num_users, self.overlapped_num_users,
self.source_user2pop, self.target_user2pop,
self.source_user_embedding.weight[:self.overlapped_num_users], self.target_user_embedding.weight)
elif self.mode == 'overlap_items':
self.build_unit_benchmark_embedding(self.total_num_items, self.overlapped_num_items,
self.source_item2pop, self.target_item2pop,
self.source_item_embedding.weight[:self.overlapped_num_items], self.target_item_embedding.weight)
def maxmin_normalize(self, embed_weight):
min_ = torch.amin(embed_weight, dim=1, keepdim=True)
max_ = torch.amax(embed_weight, dim=1, keepdim=True)
mean_ = (max_ + min_) / 2
normal_mat = (embed_weight - mean_) / (max_ - mean_)
return normal_mat, mean_, max_
def calculate_unit_map_loss(self, target_num_units, target_unit_embeddings):
sampled_index = np.random.randint(0, target_num_units, self.map_batch_size)
item_embeddings = target_unit_embeddings[sampled_index]
item_embeddings, _, _ = self.maxmin_normalize(item_embeddings)
item_embeddings = self.mapping_mlp_layers(item_embeddings)
benchmark_embeddings = self.benchmark_embedding[sampled_index]
benchmark_embeddings, _, _ = self.maxmin_normalize(benchmark_embeddings)
loss = self.map_loss(item_embeddings, benchmark_embeddings)
return loss
def calculate_map_loss(self):
loss = None
if self.mode == 'overlap_users':
loss = self.calculate_unit_map_loss(self.target_num_users, self.target_user_embedding.weight)
elif self.mode == 'overlap_items':
loss = self.calculate_unit_map_loss(self.target_num_items, self.target_item_embedding.weight)
return loss
def calculate_loss(self, interaction):
if self.phase == 'SOURCE' and self.phase2count[self.phase] == 1:
return self.calculate_rec_loss(
interaction, self.source_user_embedding.weight, self.source_item_embedding.weight,
self.SOURCE_USER_ID, self.SOURCE_ITEM_ID, self.SOURCE_NEG_ITEM_ID, self.SOURCE_LABEL)
elif self.phase == 'TARGET' and self.phase2count[self.phase] == 1:
return self.calculate_rec_loss(
interaction, self.target_user_embedding.weight, self.target_item_embedding.weight,
self.TARGET_USER_ID, self.TARGET_ITEM_ID, self.TARGET_NEG_ITEM_ID, self.TARGET_LABEL)
elif self.phase == 'BOTH':
return self.calculate_map_loss()
elif self.phase == 'TARGET' and self.phase2count[self.phase] == 2:
if self.mode == 'overlap_users':
return self.calculate_rec_loss(
interaction, self.affine_embedding, self.target_item_embedding.weight,
self.TARGET_USER_ID, self.TARGET_ITEM_ID, self.TARGET_NEG_ITEM_ID, self.TARGET_LABEL)
elif self.mode == 'overlap_items':
return self.calculate_rec_loss(
interaction, self.target_user_embedding.weight, self.affine_embedding,
self.TARGET_USER_ID, self.TARGET_ITEM_ID, self.TARGET_NEG_ITEM_ID, self.TARGET_LABEL)
def full_sort_predict(self, interaction):
user_e, all_item_e = None, None
if self.phase == 'SOURCE' and self.phase2count[self.phase] == 1:
user = interaction[self.SOURCE_USER_ID]
user_e = self.source_user_embedding(user)
all_item_e1 = self.source_item_embedding.weight[:self.overlapped_num_items]
all_item_e2 = self.source_item_embedding.weight[self.target_num_items:]
all_item_e = torch.cat([all_item_e1, all_item_e2], dim=0)
elif self.phase == 'TARGET' and self.phase2count[self.phase] == 1:
user = interaction[self.TARGET_USER_ID]
user_e = self.target_user_embedding(user)
all_item_e = self.target_item_embedding.weight[:self.target_num_items]
elif self.phase == 'TARGET' and self.phase2count[self.phase] == 2:
user = interaction[self.TARGET_USER_ID]
if self.mode == 'overlap_users':
user_e = self.affine_embedding[user]
all_item_e = self.target_item_embedding.weight[:self.target_num_items]
elif self.mode == 'overlap_items':
user_e = self.target_user_embedding(user)
all_item_e = self.affine_embedding
else:
user = interaction[self.TARGET_USER_ID]
if self.mode == 'overlap_users':
user_e = self.affine_embedding[user]
all_item_e = self.target_item_embedding.weight[:self.target_num_items]
elif self.mode == 'overlap_items':
user_e = self.target_user_embedding(user)
all_item_e = self.affine_embedding
score = torch.matmul(user_e, all_item_e.transpose(0, 1))
return score
def predict(self, interaction):
user_e, item_e = None, None
if self.phase == 'SOURCE' and self.phase2count[self.phase] == 1:
user = interaction[self.SOURCE_USER_ID]
item = interaction[self.SOURCE_ITEM_ID]
user_e = self.source_user_embedding(user)
item_e = self.source_item_embedding(item)
elif self.phase == 'TARGET' and self.phase2count[self.phase] == 1:
user = interaction[self.TARGET_USER_ID]
item = interaction[self.TARGET_ITEM_ID]
user_e = self.target_user_embedding(user)
item_e = self.target_item_embedding(item)
elif self.phase == 'TARGET' and self.phase2count[self.phase] == 2:
user = interaction[self.TARGET_USER_ID]
item = interaction[self.TARGET_ITEM_ID]
if self.mode == 'overlap_users':
user_e = self.affine_embedding[user]
item_e = self.target_item_embedding(item)
elif self.mode == 'overlap_items':
user_e = self.target_user_embedding(user)
item_e = self.affine_embedding[item]
else:
user = interaction[self.TARGET_USER_ID]
item = interaction[self.TARGET_ITEM_ID]
if self.mode == 'overlap_users':
user_e = self.affine_embedding[user]
item_e = self.target_item_embedding(item)
elif self.mode == 'overlap_items':
user_e = self.target_user_embedding(user)
item_e = self.affine_embedding[item]
score = torch.mul(user_e, item_e).sum(dim=1)
return score