-
Notifications
You must be signed in to change notification settings - Fork 0
/
generateNetwork.py
225 lines (205 loc) · 7.83 KB
/
generateNetwork.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
# -*- coding: utf-8 -*-
#==============================================================================
import torch
from torch import nn, optim
from torch.utils.data import DataLoader
from torchvision import transforms
from torchvision import datasets
from torch.autograd import Variable
import numpy as np
import librosa
#==============================================================================
learning_rate = 0.0008
num_epoches = 100
batch_size = 20 # train 235 *1000
# test row = 129761
pca_dim = 15
var_batch = 50
#========================================================================
def zeroMean(data):
meanVal = np.mean(data, axis=0)
newData = data - meanVal
return newData, meanVal
def pca(data, n = pca_dim):
newData, meanVal = zeroMean(data)
covMat = np.cov(newData, rowvar=0)
eigVals, eigVects = np.linalg.eig(np.mat(covMat))
eigValIndice = np.argsort(eigVals)
n_eigValIndice = eigValIndice[-1:-(n+1):-1]
n_eigVect = eigVects[:, n_eigValIndice]
lowDDataMat = newData * n_eigVect
reconMat = (lowDDataMat * n_eigVect.T) + meanVal
lowDDataMat = np.delete(lowDDataMat, 0, axis= 1)
reconMat = np.delete (reconMat, 0, axis=1)
return lowDDataMat, reconMat
def get_X(path):
data = np.load(path)
data = data[:,0:56]
lowDDdata, reconMat = pca(data)
var_matrix = []
dim = round(lowDDdata.shape[0]/var_batch)
batch_sc = []
for i in range(dim):
batch = lowDDdata[i*var_batch:(i+1)*var_batch]
batch_std = np.std(batch, axis = 0)
batch_var = np.var(batch, axis = 0)
batch_mean = np.mean(batch, axis = 0)
# spectral centroid
# for j in range(batch.shape[1]):
# y = np.ravel((batch[:,j]).T)
# sc = librosa.feature.spectral_centroid(y, sr=8000)[0]
# batch_sc = np.append(batch_sc,sc)
# batch_sc = np.array(batch_sc)
# batch_sc = np.reshape(batch_sc, (1, batch_sc.shape[0]))
batch_feature = np.hstack((batch_mean, batch_std, batch_var))
# batch_feature = (batch_feature[0]).tolist()
batch_feature = np.ravel(batch_feature)
var_matrix.append(batch_feature)
var_matrix = np.array(var_matrix)
var_matrix = var_matrix.reshape([dim, (pca_dim-1)*3])
return var_matrix
def get_Y(path):
label = np.load(path)
label = label.T
label = label[0]
resized_label = []
dim = round(label.size/ var_batch)
for i in range(dim):
batch_label = label[i*var_batch]
resized_label.append(batch_label)
resized_label = np.array(resized_label)
# label = torch.from_numpy(label[0])
return resized_label
def get_test_data():
data = np.load('../X_test.npy')
data = data[:,0:56]
lowDDdata, reconMat = pca(data)
var_matrix = []
label = np.load('../Y_test_label.npy')
label = label.T
label = label[0]
resized_label = []
dim = round(label.size/ var_batch)
for i in range(dim):
batch = lowDDdata[i*var_batch:(i+1)*var_batch]
batch_std = np.std(batch, axis = 0)
batch_var = np.var(batch, axis = 0)
batch_mean = np.mean(batch, axis = 0)
# spectral centroid
# for j in range(batch.shape[1]):
# y = np.ravel((batch[:,j]).T)
# sc = librosa.feature.spectral_centroid(y, sr=8000)[0]
# batch_sc = np.append(batch_sc,sc)
# batch_sc = np.array(batch_sc)
# batch_feature = np.hstack((batch_mean, batch_std, batch_var, batch_sc))
batch_feature = np.hstack((batch_mean, batch_std, batch_var))
batch_feature = np.ravel(batch_feature)
# batch_feature = (batch_feature[0]).tolist()
batch_label = label[i*var_batch:(i+1)*var_batch]
if (np.var(batch_label) == 0):
batch_label = label[i * var_batch]
var_matrix.append(batch_feature)
resized_label.append(batch_label)
resized_label = torch.from_numpy(np.array(resized_label))
var_matrix = torch.from_numpy(np.array(var_matrix))
# var_matrix = torch.tensor(var_matrix.reshape([var_matrix.shape[0], (pca_dim-1)*3]))
return var_matrix, resized_label
train_dataset = torch.from_numpy(get_X('../X_train.npy'))
train_label = torch.from_numpy(get_Y('../Y_label.npy'))
test_dataset, test_label = get_test_data()
# CUDA_LAUNCH_BLOCKING = 1
input_dim = train_dataset.size(1)
#===================================================================
# define sinple forword neural network
class Neuralnetwork(nn.Module):
def __init__(self, in_dim, nn_hidden_1, nn_hidden_2, out_dim):
super(Neuralnetwork, self).__init__()
self.layer1 = nn.Sequential(
nn.Linear(in_dim, nn_hidden_1),
nn.ReLU(True))
self.layer2 = nn.Sequential(
nn.Linear(nn_hidden_1, nn_hidden_2),
nn.Softmax()
)
# self.layer3 = nn.Sequential(
# nn.Linear(nn_hidden_2, nn_hidden_3),
# nn.ReLU(True)
# )
# self.layer4 = nn.Sequential(
# nn.Linear(nn_hidden_3, out_dim),
# nn.Softmax())
def forward(self, x):
x = self.layer1(x)
x = self.layer2(x)
# x = self.layer3(x)
# x = self.layer4(x)
return x
# ===========================================================
train_len = int(train_dataset.size(0) / batch_size)
test_len = int(test_dataset.shape[0] / batch_size)
def getbatch(index, X, Y):
start = index * batch_size
end = (index + 1) * batch_size
data = X[start:end, :]
label = Y[start:end]
return data, label
model = Neuralnetwork(1*input_dim, 80, 40, 6) # need to be modified
if torch.cuda.is_available():
model = model.cuda()
criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr = learning_rate)
for epoch in range(num_epoches):
print ('epoch{}'.format(epoch + 1))
print ('*'*10)
running_loss = 0.0
running_acc = 0.0
index = 0
for index in range(train_len):
x_train, y_train = getbatch(index, train_dataset, train_label)
if torch.cuda.is_available():
x_train = Variable(x_train).cuda()
y_train = Variable(y_train).cuda()
else :
x_train = Variable(x_train)
y_train = Variable(y_train)
# forward
out = model.forward(x_train.float())
# size1, size2 = out.size(0), out.size(1)
loss = criterion(out, y_train.long())
running_loss += loss.data * y_train.size(0)
_, pred = torch.max(out, 1)
num_correct = (pred == y_train.long()).sum()
running_acc += num_correct.data
# backward
optimizer.zero_grad()
loss.backward()
optimizer.step()
if index % 100 == 0 :
print ('{}/{} Loss: {:.6f}, Acc: {:.6f}'.format(
epoch+1, num_epoches, float(running_loss)/(batch_size*(index+1)), float(running_acc)/(batch_size * (index+1))
))
print ('Finish {} epoch, Loss: {:.6f}, Acc:{:.6f}'.format(
epoch + 1, float(running_loss) / (len(train_dataset)), float(running_acc) / (len(train_dataset))
))
model.eval()
eval_loss = 0.0
eval_acc = 0.0
for index in range(test_len):
x_test, y_test = getbatch(index, test_dataset, test_label)
if torch.cuda.is_available():
x_test = Variable(x_test).cuda()
y_test = Variable(y_test).cuda()
else :
x_test = Variable(x_test)
y_test = Variable(y_test)
out = model.forward(x_test.float())
loss = criterion(out, y_test.long())
eval_loss += loss.data * y_test.size(0)
_, pred = torch.max(out, 1)
num_correct = (pred == y_test.long()).sum()
eval_acc += num_correct.data
print ('test loss: {:.6f}, Acc: {:.6f}'.format(
float(eval_loss) / (len(test_dataset)), float(eval_acc) / (len(test_dataset)))
)
print ('Done!')
# torch.save(model.state_dict(), './model1.pth')