-
Notifications
You must be signed in to change notification settings - Fork 9
Expand file tree
/
Copy pathutils.py
More file actions
377 lines (327 loc) · 11.3 KB
/
utils.py
File metadata and controls
377 lines (327 loc) · 11.3 KB
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
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
# -*- coding: utf-8 -*-
import numpy as np
import math
import random
import os
import json
import pickle
from scipy.sparse import csr_matrix
from tqdm import tqdm
import multiprocessing
import torch
import torch.nn.functional as F
def set_seed(seed):
random.seed(seed)
os.environ['PYTHONHASHSEED'] = str(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
# some cudnn methods can be random even after fixing the seed
# unless you tell it to be deterministic
torch.backends.cudnn.deterministic = True
def check_path(path):
if not os.path.exists(path):
os.makedirs(path)
print(f'{path} created')
def neg_sample(item_set, item_size):
item = random.randint(1, item_size - 1)
while item in item_set:
item = random.randint(1, item_size - 1)
return item
class EarlyStopping:
"""Early stops the training if validation loss doesn't improve after a given patience."""
def __init__(self, checkpoint_path, patience=7, verbose=False, delta=0):
"""
Args:
patience (int): How long to wait after last time validation loss improved.
Default: 7
verbose (bool): If True, prints a message for each validation loss improvement.
Default: False
delta (float): Minimum change in the monitored quantity to qualify as an improvement.
Default: 0
"""
self.checkpoint_path = checkpoint_path
self.patience = patience
self.verbose = verbose
self.counter = 0
self.best_score = None
self.early_stop = False
self.delta = delta
def compare(self, score):
for i in range(len(score)):
if score[i] > self.best_score[i]+self.delta:
return False
return True
def __call__(self, score, model):
# score HIT@10 NDCG@10
if self.best_score is None:
self.best_score = score
self.score_min = np.array([0]*len(score))
self.save_checkpoint(score, model)
elif self.compare(score):
self.counter += 1
print(f'EarlyStopping counter: {self.counter} out of {self.patience}')
if self.counter >= self.patience:
self.early_stop = True
else:
self.best_score = score
self.save_checkpoint(score, model)
self.counter = 0
def save_checkpoint(self, score, model):
'''Saves model when validation loss decrease.'''
if self.verbose:
print(f'Validation score increased. Saving model ...')
torch.save(model.state_dict(), self.checkpoint_path)
self.score_min = score
def kmax_pooling(x, dim, k):
index = x.topk(k, dim=dim)[1].sort(dim=dim)[0]
return x.gather(dim, index).squeeze(dim)
def avg_pooling(x, dim):
return x.sum(dim=dim)/x.size(dim)
def generate_rating_matrix_valid(user_seq, num_users, num_items):
# three lists are used to construct sparse matrix
row = []
col = []
data = []
for user_id, item_list in enumerate(user_seq):
for item in item_list[:-2]: #
row.append(user_id)
col.append(item)
data.append(1)
row = np.array(row)
col = np.array(col)
data = np.array(data)
rating_matrix = csr_matrix((data, (row, col)), shape=(num_users, num_items))
return rating_matrix
def generate_rating_matrix_test(user_seq, num_users, num_items):
# three lists are used to construct sparse matrix
row = []
col = []
data = []
for user_id, item_list in enumerate(user_seq):
for item in item_list[:-1]: #
row.append(user_id)
col.append(item)
data.append(1)
row = np.array(row)
col = np.array(col)
data = np.array(data)
rating_matrix = csr_matrix((data, (row, col)), shape=(num_users, num_items))
return rating_matrix
def get_user_seqs(data_file):
lines = open(data_file).readlines()
user_seq = []
item_set = set()
for line in lines:
user, items = line.strip().split(' ', 1)
items = items.split(' ')
items = [int(item) for item in items]
user_seq.append(items)
item_set = item_set | set(items)
max_item = max(item_set)
num_users = len(lines)
num_items = max_item + 2
valid_rating_matrix = generate_rating_matrix_valid(user_seq, num_users, num_items)
test_rating_matrix = generate_rating_matrix_test(user_seq, num_users, num_items)
return user_seq, max_item, valid_rating_matrix, test_rating_matrix, num_users
def get_user_seqs_long(data_file):
lines = open(data_file).readlines()
user_seq = []
long_sequence = []
item_set = set()
for line in lines:
user, items = line.strip().split(' ', 1)
items = items.split(' ')
items = [int(item) for item in items]
long_sequence.extend(items) #
user_seq.append(items)
item_set = item_set | set(items)
max_item = max(item_set)
return user_seq, max_item, long_sequence
def get_user_seqs_and_sample(data_file, sample_file):
lines = open(data_file).readlines()
user_seq = []
item_set = set()
for line in lines:
user, items = line.strip().split(' ', 1)
items = items.split(' ')
items = [int(item) for item in items]
user_seq.append(items)
item_set = item_set | set(items)
max_item = max(item_set)
lines = open(sample_file).readlines()
sample_seq = []
for line in lines:
user, items = line.strip().split(' ', 1)
items = items.split(' ')
items = [int(item) for item in items]
sample_seq.append(items)
assert len(user_seq) == len(sample_seq)
return user_seq, max_item, sample_seq
def get_item2attribute_json(data_file):
item2attribute = json.loads(open(data_file).readline())
attribute_set = set()
for item, attributes in item2attribute.items():
attribute_set = attribute_set | set(attributes)
attribute_size = max(attribute_set) # 331
return item2attribute, attribute_size
def get_metric(pred_list, topk=10):
NDCG = 0.0
HIT = 0.0
MRR = 0.0
# [batch] the answer's rank
for rank in pred_list:
MRR += 1.0 / (rank + 1.0)
if rank < topk:
NDCG += 1.0 / np.log2(rank + 2.0)
HIT += 1.0
return HIT /len(pred_list), NDCG /len(pred_list), MRR /len(pred_list)
def precision_at_k_per_sample(actual, predicted, topk):
num_hits = 0
for place in predicted:
if place in actual:
num_hits += 1
return num_hits / (topk + 0.0)
def precision_at_k(actual, predicted, topk):
sum_precision = 0.0
num_users = len(predicted)
for i in range(num_users):
act_set = set(actual[i])
pred_set = set(predicted[i][:topk])
sum_precision += len(act_set & pred_set) / float(topk)
return sum_precision / num_users
def recall_at_k(actual, predicted, topk):
sum_recall = 0.0
num_users = len(predicted)
true_users = 0
recall_dict = {}
for i in range(num_users):
act_set = set(actual[i])
pred_set = set(predicted[i][:topk])
if len(act_set) != 0:
#sum_recall += len(act_set & pred_set) / float(len(act_set))
one_user_recall = len(act_set & pred_set) / float(len(act_set))
recall_dict[i] = one_user_recall
sum_recall += one_user_recall
true_users += 1
return sum_recall / true_users, recall_dict
def cal_mrr(actual, predicted):
sum_mrr = 0.
true_users = 0
num_users = len(predicted)
mrr_dict = {}
for i in range(num_users):
r = []
act_set = set(actual[i])
pred_list = predicted[i]
for item in pred_list:
if item in act_set:
r.append(1)
else:
r.append(0)
r = np.array(r)
if np.sum(r) > 0:
#sum_mrr += np.reciprocal(np.where(r==1)[0]+1, dtype=np.float)[0]
one_user_mrr = np.reciprocal(np.where(r==1)[0]+1, dtype=np.float)[0]
sum_mrr += one_user_mrr
true_users += 1
mrr_dict[i] = one_user_mrr
else:
mrr_dict[i] = 0.
return sum_mrr / len(predicted), mrr_dict
def apk(actual, predicted, k=10):
"""
Computes the average precision at k.
This function computes the average precision at k between two lists of
items.
Parameters
----------
actual : list
A list of elements that are to be predicted (order doesn't matter)
predicted : list
A list of predicted elements (order does matter)
k : int, optional
The maximum number of predicted elements
Returns
-------
score : double
The average precision at k over the input lists
"""
if len(predicted)>k:
predicted = predicted[:k]
score = 0.0
num_hits = 0.0
for i,p in enumerate(predicted):
if p in actual and p not in predicted[:i]:
num_hits += 1.0
score += num_hits / (i+1.0)
if not actual:
return 0.0
return score / min(len(actual), k)
def mapk(actual, predicted, k=10):
"""
Computes the mean average precision at k.
This function computes the mean average prescision at k between two lists
of lists of items.
Parameters
----------
actual : list
A list of lists of elements that are to be predicted
(order doesn't matter in the lists)
predicted : list
A list of lists of predicted elements
(order matters in the lists)
k : int, optional
The maximum number of predicted elements
Returns
-------
score : double
The mean average precision at k over the input lists
"""
return np.mean([apk(a, p, k) for a, p in zip(actual, predicted)])
def ndcg_k(actual, predicted, topk):
res = 0
ndcg_dict = {}
for user_id in range(len(actual)):
k = min(topk, len(actual[user_id]))
idcg = idcg_k(k)
dcg_k = sum([int(predicted[user_id][j] in
set(actual[user_id])) / math.log(j+2, 2) for j in range(topk)])
res += dcg_k / idcg
ndcg_dict[user_id] = dcg_k / idcg
return res / float(len(actual)), ndcg_dict
# Calculates the ideal discounted cumulative gain at k
def idcg_k(k):
res = sum([1.0/math.log(i+2, 2) for i in range(k)])
if not res:
return 1.0
else:
return res
def dcg_at_k(r, k, method=1):
"""Score is discounted cumulative gain (dcg)
Relevance is positive real values. Can use binary
as the previous methods.
Returns:
Discounted cumulative gain
"""
r = np.asfarray(r)[:k]
if r.size:
if method == 0:
return r[0] + np.sum(r[1:] / np.log2(np.arange(2, r.size + 1)))
elif method == 1:
return np.sum(r / np.log2(np.arange(2, r.size + 2)))
else:
raise ValueError('method must be 0 or 1.')
return 0.
def ndcg_at_k(r, k, method=1):
"""Score is normalized discounted cumulative gain (ndcg)
Relevance is positive real values. Can use binary
as the previous methods.
Returns:
Normalized discounted cumulative gain
"""
dcg_max = dcg_at_k(sorted(r, reverse=True), k, method)
if not dcg_max:
return 0.
return dcg_at_k(r, k, method) / dcg_max