-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathSepPCNET.py
168 lines (150 loc) · 7.38 KB
/
SepPCNET.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
import torch
import torch.utils.data as Data
from torch.utils.data import Dataset
import numpy as np
from torch import nn, optim
import torch.nn.functional as F
from torch.utils.data.sampler import WeightedRandomSampler
import sklearn
from sklearn import metrics
import sys
torch.manual_seed(42)
coordinate_collect_train=[]
label_collect_train=[]
with open('./trainset.txt',encoding='utf-8') as f:
for line in f:
line.rstrip()
item=line.split()
cas=item[0]
label=item[1]
coordinate=[]
with open('./dataset/'+cas+'sparse.txt',encoding='utf-8') as h:
for hang in h:
hang.rstrip()
hangproper=hang.split()
coordinate.append((float(hangproper[0]),float(hangproper[1]),float(hangproper[2]),float(hangproper[3])))
coordinate_collect_train.append(coordinate)
label_collect_train.append(int(label))
coordinate_collect_train=np.array(coordinate_collect_train)
np.savez_compressed('./trainset.npz', arr=coordinate_collect_train)
label_collect_train=np.array(label_collect_train)
np.savez_compressed('./trainset_label.npz', arr=label_collect_train)
#loading .npz file directly helps to save time
coordinate=np.load('./trainset.npz')['arr'].astype(np.float32)
print(coordinate.shape)
label=np.load('./trainset_label.npz')['arr'].astype(np.float32)
train_coordinate=torch.from_numpy(coordinate)
train_coordinate=train_coordinate.permute(0,2,1)
train_label=torch.Tensor(label)
trainset=Data.TensorDataset(train_coordinate,train_label)
BATCH_SIZE_train=64
weights=[8 if label==1 else 1 for coordinate, label in trainset]
sampler=WeightedRandomSampler(weights,num_samples=1920,replacement=True)
trainset_loader=Data.DataLoader(trainset,batch_size=BATCH_SIZE_train,sampler=sampler,drop_last=True)
coordinate_collect_validation=[]
label_collect_validation=[]
with open('./validationset.txt',encoding='utf-8') as f:
for line in f:
line.rstrip()
item=line.split()
cas=item[0]
label=item[1]
coordinate=[]
with open('./dataset/'+cas+'sparse.txt',encoding='utf-8') as h:
for hang in h:
hang.rstrip()
hangproper=hang.split()
coordinate.append((float(hangproper[0]),float(hangproper[1]),float(hangproper[2]),float(hangproper[3])))
coordinate_collect_validation.append(coordinate)
label_collect_validation.append(int(label))
coordinate_collect_validation=np.array(coordinate_collect_validation)
np.savez_compressed('./validationset.npz', arr=coordinate_collect_validation)
label_collect_validation=np.array(label_collect_validation)
np.savez_compressed('./validationset_label.npz', arr=label_collect_validation)
coordinate=np.load('./validationset.npz')['arr'].astype(np.float32)
print(coordinate.shape)
label=np.load('./validationset_label.npz')['arr'].astype(np.float32)
validation_coordinate=torch.from_numpy(coordinate)
validation_coordinate=validation_coordinate.permute(0,2,1)
validation_label=torch.Tensor(label)
class CoordinatetoPredict(nn.Module):
def __init__(self):
super(CoordinatetoPredict,self).__init__()
self.conv1=nn.Sequential(nn.Conv1d(4,64,kernel_size=1),nn.BatchNorm1d(64),nn.ReLU(),
nn.Conv1d(64,64,kernel_size=1),nn.BatchNorm1d(64),nn.ReLU(),
nn.Conv1d(64,128,kernel_size=1),nn.BatchNorm1d(128),nn.ReLU(),
nn.Conv1d(128,1024,kernel_size=1),nn.BatchNorm1d(1024),nn.ReLU(),
nn.MaxPool1d(kernel_size=4096))
self.fc2=nn.Sequential(nn.Linear(1024,256),nn.BatchNorm1d(256),nn.ReLU(),
nn.Linear(256,64),nn.BatchNorm1d(64),nn.ReLU(),
nn.Linear(64,8),nn.BatchNorm1d(8),nn.ReLU(),
nn.Linear(8,1),nn.Sigmoid())
def forward(self, x):
#print(x.shape)
x_conv1=self.conv1(x)
#print(x_conv1.shape)
x_conv1=x_conv1.view(-1,x_conv1.size(1))
#print(x_conv1.shape)
x_fc2=self.fc2(x_conv1)
#print(x_fc2.shape)
x_out=x_fc2.view(x_fc2.size(0))
return x_out
L2=0.001
learning_rate=0.001
model=CoordinatetoPredict()
optimizer=optim.Adam(model.parameters(),weight_decay=L2,lr=learning_rate)
criterion=nn.BCELoss()
with open ('./conv_structure1/minibatch64/{}L2/lr{}/result/trainres.txt'.format(L2,learning_rate),mode='w',encoding="utf-8") as h:
print('Epoch\tLoss\tAUC',file=h)
with open ('./conv_structure1/minibatch64/{}L2/lr{}/result/validationres.txt'.format(L2,learning_rate),mode='w',encoding="utf-8") as f:
print('Epoch\tLoss\tAUC',file=f)
AUC_validation_max=0.5
consecutiveepoch_num=0
for epoch in range(60):
loss_epoch=0
model.train()
for batch_data_train in trainset_loader:
batch_coordinate_train, batch_label_train=batch_data_train
print(batch_coordinate_train.shape)
out=model(batch_coordinate_train)
batch_label_train=batch_label_train.float()
loss_train=criterion(out,batch_label_train)
batch_loss_train=loss_train.item()
loss_epoch=loss_epoch+batch_loss_train*batch_label_train.size(0)
out_numpy=out.detach().numpy()
batchlabel_numpy_train=batch_label_train.detach().numpy()
AUC_train=metrics.roc_auc_score(batchlabel_numpy_train,out_numpy)
optimizer.zero_grad()
loss_train.backward()
optimizer.step()
if (epoch+1) %1 ==0:
print('*'*10)
print('epoch {}, Loss {}'.format((epoch+1),(loss_epoch/1920)))
print('AUC of Training set: {}'.format(AUC_train))
with open('./conv_structure1/minibatch64/{}L2/lr{}/result/trainres.txt'.format(L2,learning_rate),mode='a',encoding="utf-8") as h:
print(str(epoch+1)+'\t'+str(loss_epoch/1920)+'\t'+str(AUC_train),file=h)
model.eval()
out=model(validation_coordinate)
validation_label=validation_label.float()
loss_validation=criterion(out,validation_label)
batch_loss_validation=loss_validation.item()
out_numpy=out.detach().numpy()
batchlabel_numpy_validation=validation_label.detach().numpy()
AUC_validation=metrics.roc_auc_score(batchlabel_numpy_validation,out_numpy)
print('*'*10)
print('epoch {}, Loss {}'.format((epoch+1),(batch_loss_validation)))
print('AUC of Validation set: {}'.format(AUC_validation))
if (epoch+1) % 1 ==0:
with open('./conv_structure1/minibatch64/{}L2/lr{}/result/validationres.txt'.format(L2,learning_rate),mode='a',encoding="utf-8") as f:
print(str(epoch+1)+'\t'+str(batch_loss_validation)+'\t'+str(AUC_validation),file=f)
if (epoch+1) % 1 ==0:
torch.save(model.state_dict(),'./conv_structure1/minibatch64/{}L2/lr{}/ref/train-{:03d}.pth'.format(L2,learning_rate,epoch+1))
if AUC_validation > AUC_validation_max:
AUC_validation_max=AUC_validation
consecutiveepoch_num=0
print('{} consecutive epoches without AUC increase'.format(consecutiveepoch_num))
else:
consecutiveepoch_num+=1
print('{} consecutive epoches without AUC increase'.format(consecutiveepoch_num))
if consecutiveepoch_num>=15:
sys.exit(0)