-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathageGenderClassification.py
264 lines (232 loc) · 10 KB
/
ageGenderClassification.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
#!/usr/bin/env python
# -*- coding:utf-8 -*-
import numpy as np
import pandas as pd
from multiprocessing import Pool, Manager
import copy
import torch as tc
import torch.utils.data as Data
from torch.autograd import Variable as var
import Utils as ut
logEpochFlag="==============================================================================\n"
root = "./statisticLearning/proj3_data/task1/"
moviesGenres = {'Action': 0,
'Adventure': 1,
'Animation': 2,
'Children\'s': 3,
'Comedy': 4,
'Crime': 5,
'Documentary': 6,
'Drama': 7,
'Fantasy': 8,
'Film-Noir': 9,
'Horror': 10,
'Musical': 11,
'Mystery': 12,
'Romance': 13,
'Sci-Fi': 14,
'Thriller': 15,
'War': 16,
'Western': 17}
usersGenders = {'M': 0, 'F': 1}
usersAges = [1, 18, 25, 35, 45, 50, 56]
LR = 0.1
BATCH_SIZE = [32,5]
EPOCH = 1
class AgeGenderClassification(object):
def __init__(self):
super(AgeGenderClassification, self).__init__()
self.movies = None
self.ratings = None
self.users = None # 604 rows per file
# because there has 18 classes in movie genres,genderNn is a list which has 18 little classifier correspondingly
self.genderNn = [NetGender(1,2)] * 18
self.ageNn = [NetAge(1,7)] * 18
# with ut.Log(logEpochFlag+"netGender:"+str(self.genderNn[0])+"\n"+"netAge:"+str(self.ageNn[0])+"\n"):
# pass
self.optimizer = [tc.optim.Adam(i.parameters(), lr=LR, betas=(0.9, 0.99)) for i in self.genderNn]
self.loss = [tc.nn.CrossEntropyLoss()] * 18
def moviesGenresPreprocess(self, x):
tmp = x[-1].split("|")
x[-1] = [moviesGenres[i] for i in tmp]
return x
def ratingsPreprocess(self, x):
tmp = self.movies[x[1]]
for i in tmp:
self.tmp.append(np.array([x[0], i, x[2]]).reshape(1, 3))
return x
def genderAndAgeClassifier(self, train, test,isGender=True):
error, length = 0, 0
if isGender:
bSize=BATCH_SIZE[0]
else:
bSize=BATCH_SIZE[1]
# print(logEpochFlag)
for i in range(18): # train 18 classifier respectively
if test.get(i) == None: # no test data,so that we needn't train this network
continue
length += len(test[i])
if train.get(i) != None: # avoid problem that some classes may have no data
test_data = np.concatenate(test[i])
test_x = var(tc.unsqueeze((tc.from_numpy(test_data[:, -1]).type(tc.FloatTensor) - 1) / 4,dim=1),volatile=True)
test_y = tc.from_numpy(test_data[:, 1]).type(tc.LongTensor)
data = np.concatenate(train[i])
dataSet = Data.TensorDataset(data_tensor=tc.unsqueeze((tc.from_numpy(data[:, -1]).type(tc.FloatTensor) - 1) / 4,dim=1),
target_tensor=tc.from_numpy(data[:, 1]).type(tc.LongTensor))
loader = Data.DataLoader(dataset=dataSet, batch_size=bSize, shuffle=True, num_workers=10)
for epoch in range(EPOCH):
for step, (x, y) in enumerate(loader):
_x = var(x)
_y = var(y)
if isGender:
output = self.genderNn[i](_x)
else:
output = self.ageNn[i](_x)
loss = self.loss[i](output, _y)
self.optimizer[i].zero_grad()
loss.backward()
self.optimizer[i].step()
pred_y = tc.max(self.genderNn[i](test_x), 1)[1].data.squeeze()
if isGender:
error1 = sum(pred_y != test_y)
if error1 / len(test[i])<0.37: # it means that train data is efficient,unless do not predict test data
error += error1
else:
for i in zip(pred_y,test_y):
error1=abs(i[0] - i[1])
if error1<3: # it means that train data is efficient,unless do not predict test data
error+=error1
else: # for some classes,train sets are null,but test sets are not null
# error += len(test[i]) / 2
pass
return error / length
def getData(self):
self.movies = pd.read_csv(root + "movies.dat", sep='::', engine='python', names=['mID', 'mTitle', 'mGenre'],
dtype={'mID': int, 'mTitle': str, 'mGenre': str}).dropna(0, 'all').as_matrix()
self.movies = self.movies[:, [0, 2]]
list(map(self.moviesGenresPreprocess, self.movies))
self.movies = dict(zip(self.movies[:, 0], self.movies[:, 1]))
self.ratings = pd.read_csv(root + "ratings.dat", sep='::', engine='python', names=['uID', 'mID', 'rating'],
usecols=[0, 1, 2],
dtype=int).dropna(0, 'all').as_matrix()
self.tmp = []
list(map(self.ratingsPreprocess, self.ratings))
self.ratings = np.concatenate(self.tmp)
del self.tmp
# with ut.Timing("read data cost"):
self.users = multiReadProc(list(range(10)))
def genderAndAgeCrossValidation(self,isGender=True):
# change ratings's userID to userGender,and split test data and train data
# then 10-cross validation
if isGender:
index=1
else:
index=2
self.tmp = dict(zip(self.ratings[:, 0], self.ratings[:, 1:]))
errorRate = [0] * 10
for i in range(10):
users = copy.deepcopy(self.users)
test = users.pop(i)
train = np.concatenate(users)
self.tmp1 = {}
for x in test:
key = self.tmp[x[0]][0]
value3 = self.tmp[x[0]][1]
if self.tmp1.get(key) != None:
self.tmp1[key].append(np.array([x[0], x[index], value3]).reshape(1, 3))
else:
self.tmp1[key] = [np.array([x[0], x[index], value3]).reshape(1, 3)]
test = self.tmp1
self.tmp1 = {}
for x in train:
key = self.tmp[x[0]][0]
value3 = self.tmp[x[0]][1]
if self.tmp1.get(key) != None:
self.tmp1[key].append(np.array([x[0], x[index], value3]).reshape(1, 3))
else:
self.tmp1[key] = [np.array([x[0], x[index], value3]).reshape(1, 3)]
train = self.tmp1
if isGender:
# with ut.Timing("genderClass, cross"+str(i)):
errorRate[i] = self.genderAndAgeClassifier(train, test)
else:
# with ut.Timing("ageClass, cross" + str(i)):
errorRate[i] = self.genderAndAgeClassifier(train, test,False)
del self.tmp1, self.tmp
return errorRate
class NetGender(tc.nn.Module):
def __init__(self, n_input, n_output, n_hidden=10):
super(NetGender, self).__init__()
self.hidden = tc.nn.Linear(n_input, n_hidden)
self.activate = tc.nn.ReLU()
self.out = tc.nn.Linear(n_hidden, n_output)
# self.drop=tc.nn.Dropout(0.5)
tc.nn.init.normal(self.hidden.weight, 0, 0.001) # there cannot use parameters() function to init the parameters
tc.nn.init.normal(self.out.weight, 0, 0.01)
def forward(self, x):
x = self.hidden(x)
# x = self.drop(x)
x = self.activate(x)
x = self.out(x)
return x
class NetAge(tc.nn.Module):
def __init__(self, n_input, n_output, n_hidden=64):
super(NetAge, self).__init__()
self.hidden = tc.nn.Linear(n_input, n_hidden)
self.activate = tc.nn.ReLU()
self.out = tc.nn.Linear(n_hidden, n_output)
# self.drop=tc.nn.Dropout(0.5)
tc.nn.init.normal(self.hidden.weight, 0, 0.001) # there cannot use parameters() function to init the parameters
tc.nn.init.normal(self.out.weight, 0, 0.01)
def forward(self, x):
x = self.hidden(x)
# x = self.drop(x)
x = self.activate(x)
x = self.out(x)
return x
def proc(i):
tmp = pd.read_csv(root + "users.dat" + str(i), sep='::', engine='python',
names=['uID', 'uGender', 'uAge'],
dtype={'uID': int, 'uGender': str, 'uAge': int}, usecols=[0, 1, 2]).dropna(0,
'all').as_matrix()
list(map(usersGendersAndAgesPreprocess, tmp))
with lock:
data.append(tmp)
def usersGendersAndAgesPreprocess(x):
x[1] = usersGenders[x[1]]
for j, i in enumerate(usersAges):
if x[2] == i:
x[2] = j
break
return x
def multiReadProc(num):
manager = Manager()
data = manager.list()
lock = manager.Lock()
pool = Pool(initializer=globalVarinit, initargs=(lock, data)) # default number of processes is os.cpu_count()
pool.map(proc, num)
pool.close()
pool.join()
return data
def globalVarinit(_lock, _data):
global data
global lock
data = _data
lock = _lock
# @ut.timing("program run")
def main():
cla = AgeGenderClassification()
cla.getData()
error = cla.genderAndAgeCrossValidation()
print("10 fold error:" + str(error) + "\n")
averageErr = sum(error) / 10
print("averageErr:" + str(averageErr) + "\n")
# with ut.Log(logEpochFlag+"10 fold error:" + str(error) + "\n"+"averageErr:" + str(averageErr) + "\n"):
# pass
error = cla.genderAndAgeCrossValidation(isGender=False)
print("10 fold error:" + str(error) + "\n")
averageErr = sum(error) / 10
print("averageErr:" + str(averageErr) + "\n")
# with ut.Log(logEpochFlag+"10 fold error:" + str(error) + "\n"+"averageErr:" + str(averageErr) + "\n"):
# pass
main()