FilesExpand file tree

main

/

embed_creators.py

Copy path

Blame

More file actions

Blame

More file actions

Latest commit

 

History

History

History

57 lines (44 loc) · 2.45 KB

main

/

embed_creators.py

Copy path

Top

File metadata and controls

• Code
• Blame

57 lines (44 loc) · 2.45 KB

Raw

Copy raw file

Download raw file

Open symbols panel

Edit and raw actions

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

import numpy as np

DATA_DIR = './Data/Adressa'

SAVE_DIR = './SimuLine/Preprocessing/Data'

def softmax(f):

# instead: first shift the values of f so that the highest number is 0:

f -= np.max(f) # f becomes [-666, -333, 0]

return np.exp(f) / np.sum(np.exp(f)) # safe to do, gives the correct answer

def load_raw_data():

# user_article.txt, creator_article.txt, article_encode.npy

with open(DATA_DIR + '/user_article_5_hot.txt', 'r') as f:

user_article_str = f.read()

user_article = user_article_str.split('\n')

user_article = [record.split('\t') for record in user_article]

user_article = [[int(record[0]), int(record[1])] for record in user_article]

with open(DATA_DIR + '/creator_article_5_hot.txt', 'r') as f:

creator_article_str = f.read()

creator_article = creator_article_str.split('\n')

creator_article = [record.split('\t') for record in creator_article]

creator_article = [[int(record[0]), int(record[1])] for record in creator_article]

article_encode = np.load(DATA_DIR + '/article_encode_5_hot.npy')

return user_article, creator_article, article_encode

def main():

user_article, creator_article, article_encode = load_raw_data()

# Count Degrees

user_article_numpy = np.array(user_article)

creator_article_numpy = np.array(creator_article)

article_degrees = np.zeros([article_encode.shape[0]])

for record in user_article_numpy:

article_degrees[record[1]] += 1

creator_article_encode = [[] for i in range(creator_article_numpy[:, 0].max() + 1)]

creator_article_weight = [[] for i in range(creator_article_numpy[:, 0].max() + 1)]

for record in creator_article_numpy:

creator_article_encode[record[0]].append(article_encode[record[1]])

creator_article_weight[record[0]].append(article_degrees[record[1]])

creator_embedding = np.zeros([creator_article_numpy[:, 0].max() + 1, 100])

for i in range(len(creator_embedding)):

encode_seq = np.array(creator_article_encode[i])

if len(encode_seq.shape) == 1:

encode_seq = np.expand_dims(encode_seq, 0)

weight_seq = np.array(creator_article_weight[i])

weight_seq = softmax(weight_seq)

weight_seq = np.expand_dims(weight_seq, 0)

creator_embedding[i] = np.matmul(weight_seq, encode_seq)

np.save('{}/creator_embedding.npy'.format(SAVE_DIR), creator_embedding)