-
Notifications
You must be signed in to change notification settings - Fork 2
/
HamHD-ytb.py
142 lines (116 loc) · 3.82 KB
/
HamHD-ytb.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
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder
from sklearn.metrics.pairwise import cosine_similarity
from sklearn.metrics import accuracy_score
f_ = './ytb_all.csv'
def ReadCSV(f_path):
df = pd.read_csv(f_path, delimiter=',', encoding='latin-1')
X = df.CONTENT
Y = df.CLASS
le = LabelEncoder()
Y = le.fit_transform(Y)
return X, Y
def ReadCSV_(f_path):
df = pd.read_csv(f_, delimiter=',', encoding='latin-1')
text_messages = df.CONTENT
classes = df.CLASS
le = LabelEncoder()
classes = le.fit_transform(classes)
return text_messages, classes
def memGen(dim=10000, num_char=37):
dictMem = np.random.randint(2, size=(num_char, dim), dtype='int32')
dictMem[dictMem == 0] = -1
return dictMem
def encode(msg, dictMem, dim=10000):
HV = np.zeros(dim, dtype='int32')
letter_idx = 0
for letter in msg:
if letter == -1:
np.roll(HV, 500)
else:
HV = np.add(HV, dictMem[letter])
letter_idx += 1
HV_avg = np.average(HV)
HV[HV > HV_avg] = 1
HV[HV < HV_avg] = -1
HV[HV == HV_avg] = 0
return HV
def train(X, Y, dictMem, dim, n_class, alpha):
refMem = np.zeros((n_class, dim), dtype='int32')
msg_idx = 0
for msg in X:
HV = encode(msg, dictMem, dim)
refMem[Y[msg_idx]] = np.add(refMem[Y[msg_idx]], alpha * HV)
msg_idx += 1
return refMem
def test(X, Y, dictMem, refMem, dim=10000):
Y_pred = []
msg_idx = 0
for msg in X:
HV = encode(msg, dictMem, dim=dim)
sim = [0, 0]
sim[0] = cosine_similarity([HV, refMem[0]])[0][1]
sim[1] = cosine_similarity([HV, refMem[1]])[0][1]
if sim[0] > sim[1]:
Y_pred.append(0)
else:
Y_pred.append(1)
msg_idx += 1
return accuracy_score(Y, Y_pred)
def retrain(X, Y, dictMem, refMem, dim, alpha):
msg_idx = 0
for msg in X:
HV = encode(msg, dictMem, dim)
sim = [0, 0]
sim[0] = cosine_similarity([HV, refMem[0]])[0][1]
sim[1] = cosine_similarity([HV, refMem[1]])[0][1]
if sim[0] > sim[1]:
y_pred = 0
else:
y_pred = 1
if y_pred != Y[msg_idx]:
refMem[Y[msg_idx]] = np.add(refMem[Y[msg_idx]], alpha * HV)
refMem[y_pred] = np.subtract(refMem[y_pred], alpha * HV)
msg_idx += 1
return refMem
dim = 10000
num_char = 37
num_epoch = 20
num_class = 2
downsample = None
if __name__ == '__main__':
X, Y = ReadCSV(f_)
X_ = []
#print(len(X))
for item in X:
token_item = []
for letter in item.lower():
# print(letter)
if ord(letter) >= ord('a') and ord(letter) <= ord('z'):
token_item.append(ord(letter) - ord('a') + 11)
elif ord(letter) >= ord('0') and ord(letter) <= ord('9'):
token_item.append(ord(letter) - ord('0') + 1)
elif letter == ' ':
token_item.append(-1)
else:
pass
#token_item.append(0)
X_.append(token_item)
X_train, X_test, Y_train, Y_test = train_test_split(
X_, Y, test_size=0.1, random_state=19720)
X_train = X_train[:downsample]
Y_train = Y_train[:downsample]
dictMem = memGen(dim=dim, num_char=num_char)
# print(dictMem.shape)
# input()
refMem = train(X_train, Y_train, dictMem, dim, num_class, alpha = num_epoch)
print(cosine_similarity([refMem[0], refMem[1]]))
acc = test(X_test, Y_test, dictMem, refMem, dim)
print(acc)
for epoch in range(num_epoch):
refMem = retrain(X_train, Y_train, dictMem, refMem, dim, alpha = num_epoch - epoch)
acc = test(X_test, Y_test, dictMem, refMem, dim)
print(acc)