-
Notifications
You must be signed in to change notification settings - Fork 11
/
main_DBN_new.py
183 lines (152 loc) · 7.28 KB
/
main_DBN_new.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
#!/usr/bin/env python
"""
An example of running multilayer perceptrons.
Yifeng Li
CMMT, UBC, Vancouver
Sep. 23, 2014
Contact: yifeng.li.cn@gmail.com
"""
#qsub -l procs=1,pmem=2000mb,walltime=12:00:00 -r n -N main_train_test_cv -o main_train_test_cv.out -e main_train_test_cv.err -M yifeng.li.cn@gmail.com -m bea main_DBN.py
import os
#os.environ['THEANO_FLAGS']='device=cpu,base_compile=/var/tmp'
import sys
import numpy
import DBN
import classification as cl
from gc import collect as gc_collect
numpy.warnings.filterwarnings('ignore') # Theano causes some warnings
# taking the input parameters
#cell=sys.argv[1] # cell type
#wid=sys.argv[2] # window size
path="/home/yifengli/prog/my/deep_learning_v1_1/"
os.chdir(path)
"""
A data set includes three files:
[1]. A TAB seperated txt file, each row is a sample, each column is a feature.
No row and columns allowd in the txt file.
If an original sample is a matrix (3-way array), a row of this file is actually a vectorized sample,
by concatnating the rows of the original sample.
[2]. A txt file including the class labels.
Each row is a string (white space not allowed) as the class label of the corresponding row in [1].
[3]. A txt file including the name of features.
Each row is a string (white space not allowed) as the feature name of the corresponding column in [1].
"""
data_dir="/home/yifengli/prog/my/deep_learning_v1_1/data/"
result_dir="/home/yifengli/prog/my/deep_learning_v1_1/result/"
#cells=["GM12878","HepG2","K562","HelaS3","HUVEC","A549","MCF7","HMEC"]
#wids=[200,500,1000,2000,4000]
cells=["GM12878"]
wids=[200]
for cell in cells:
for wid in wids:
filename=data_dir + cell + "_" + str(wid) + "bp_Data.txt";
data=numpy.loadtxt(filename,delimiter='\t',dtype='float32')
filename=data_dir + cell + "_" + str(wid) + "bp_Classes.txt";
classes=numpy.loadtxt(filename,delimiter='\t',dtype=object)
filename=data_dir+ cell + "_Features.txt"
features=numpy.loadtxt(filename,delimiter='\t',dtype=object)
given=["A-E","I-E","A-P","I-P","A-X","I-X","UK"]
#given=["A-E","I-E"]
#given=["A-P","I-P"]
#given=["A-E","A-P"]
#given=["A-E","A-X"]
#given=["A-P","A-X"]
#given=["A-E","A-P","A-X"]
#given=["A-E","I-E","A-P","I-P"]
#given=["A-E","I-E","A-P","I-P","A-X","I-X"]
#given=["I-E","I-P"]
data,classes,_=cl.take_some_classes(data,classes,given=given,others=None)
# balance the sample sizes of the classes
rng=numpy.random.RandomState(1000)
data,classes,others=cl.balance_sample_size(data,classes,others=None,min_size_given=None,rng=rng)
print data.shape
print numpy.unique(classes)
#group=[["A-E"],["I-E"],["A-P"],["I-P"],["A-X"],["I-X"],["UK"]]
#group=[["A-E","A-P"],["I-E","I-P","A-X","I-X","UK"]]
#group=[["A-E","A-P","A-X"],["I-E","I-P","I-X","UK"]]
group=[["A-E"],["A-P"],["I-E","I-P","A-X","I-X","UK"]]
#group=[["A-E"],["A-P"],["A-X"],["I-E","I-P","I-X","UK"]]
#group=[["A-E"],["I-E"]]
#group=[["A-P"],["I-P"]]
#group=[["A-E"],["A-P"]]
#group=[["A-E"],["A-X"]]
#group=[["A-P"],["A-X"]]
#group=[["A-E"],["A-P"],["A-X"]]
#group=[["A-E","I-E"],["A-P","I-P"]]
#group=[["A-E","A-P"],["I-E","I-P"]]
#group=[["A-E","I-E"],["A-P","I-P"],["A-X","I-X"]]
#group=[["A-E","A-P","A-X"],["I-E","I-P","I-X"]]
#group=[["I-E"],["I-P"]]
classes=cl.merge_class_labels(classes,group)
print numpy.unique(classes)
classes_unique,classes=cl.change_class_labels(classes)
print numpy.unique(classes)
# set random state
rng=numpy.random.RandomState(2000)
data,classes,others=cl.balance_sample_size(data,classes,others=None,min_size_given=None,rng=rng)
# permute data to speed up learning
data_permute_id=rng.permutation(len(data))
data=data[data_permute_id,:]
classes=classes[data_permute_id]
print data.shape
print numpy.unique(classes)
kfolds=10
ind_folds=cl.kfold_cross_validation(classes,k=kfolds,shuffle=True,rng=rng)
for i in range(kfolds):
test_set_x_org=data[ind_folds==i,:]
test_set_y_org=classes[ind_folds==i]
train_set_x_org,train_set_y_org,valid_set_x_org,valid_set_y_org,_,_=cl.partition_train_valid_test(data[ind_folds!=i,:],classes[ind_folds!=i],ratio=(3,1,0),rng=rng)
# normalization
train_set_x_org,data_min,data_max=cl.normalize_col_scale01(train_set_x_org,tol=1e-10)
valid_set_x_org,_,_=cl.normalize_col_scale01(valid_set_x_org,tol=1e-10,data_min=data_min,data_max=data_max)
test_set_x_org,_,_=cl.normalize_col_scale01(test_set_x_org,tol=1e-10,data_min=data_min,data_max=data_max)
# setting the parameter
pretrain_lr=0.01
finetune_lr=0.1
alpha=0.1
lambda_reg=0#0.0001
alpha_reg=0#0.5
n_hidden=[256,64,16]
persistent_k=15
pretraining_epochs=5
training_epochs=100
batch_size=100
# train
classifier,training_time=DBN.train_model(train_set_x_org=train_set_x_org, train_set_y_org=train_set_y_org,
valid_set_x_org=valid_set_x_org, valid_set_y_org=valid_set_y_org,
pretrain_lr=pretrain_lr,finetune_lr=finetune_lr, alpha=alpha,
lambda_reg=lambda_reg, alpha_reg=alpha_reg,
n_hidden=n_hidden, persistent_k=persistent_k,
pretraining_epochs=pretraining_epochs, training_epochs=training_epochs,
batch_size=batch_size, rng=rng)
# test
test_set_y_pred,test_set_y_pred_prob,test_time=DBN.test_model(classifier, test_set_x_org, batch_size=100)
# evaluate classification performance
perf_i,conf_mat_i=cl.perform(test_set_y_org,test_set_y_pred,numpy.unique(train_set_y_org))
print perf_i
print conf_mat_i
if i==0:
perf=perf_i
conf_mat=conf_mat_i
training_times=training_time
test_times=test_time
else:
perf=numpy.vstack((perf,perf_i))
conf_mat=conf_mat+conf_mat_i
training_times=training_times + training_time
test_times=test_times + test_time
# calculate mean performance and std
perf_mean=numpy.mean(perf,axis=0)
perf_std=numpy.std(perf,axis=0)
print perf_mean
print perf_std
print conf_mat
# save the performance
save_dir=result_dir + "_".join(classes_unique)
try:
os.makedirs(save_dir)
except OSError:
pass
filename=cell + "_" + str(wid) + "bp.txt"
cl.save_perform(save_dir,filename,perf=perf_mean,std=perf_std,conf_mat=conf_mat,classes_unique=classes_unique,training_time=training_times,test_time=test_times)
gc_collect()