-
Notifications
You must be signed in to change notification settings - Fork 0
/
cnn_lasagne.py
140 lines (110 loc) · 4.31 KB
/
cnn_lasagne.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
#!/usr/bin/env python2
# -*- coding: utf-8 -*-
"""
Created on Tue Feb 28 18:04:05 2017
@author: bychkov
"""
import time
import numpy as np
import lasagne
import lasagne.layers as ll
import lasagne.nonlinearities as ln
from lasagne.objectives import binary_crossentropy
import theano
import theano.tensor as T
import matplotlib.pyplot as plt
import nolearn.lasagne as nolas
import myutils
#----------------------------------------------------------------------
# Settings:
#----------------------------------------------------------------------
data = np.load('synthetic_imgs/first.npz')
class_names = np.array(['benign','cancer'])
train_x = data['train_x']
train_y = data['train_y']
test_x = data['test_x']
test_y = data['test_y']
#Xs = np.concatenate((test_x, train_x), axis=0)
#Ys = np.concatenate((test_y, train_y), axis=0)
#----------------------------------------------------------------------
# Visualize some data:
#----------------------------------------------------------------------
plt.figure(figsize=[6,5])
for i in range(4):
plt.subplot(2,2,i+1)
plt.xlabel(class_names[train_y[i]])
plt.imshow(np.transpose(train_x[i],[1,2,0]))
plt.close()
del i
#----------------------------------------------------------------------
# Define network architecture:
#----------------------------------------------------------------------
input_var = T.tensor4('inputs')
target_var = T.ivector('targets')
#--- Network --- #
net = ll.InputLayer( shape=(None, 3, 90, 90), input_var=input_var )
net = ll.Conv2DLayer( net, 32, (3,3), nonlinearity=ln.sigmoid )
net = ll.Pool2DLayer( net, pool_size=(2,2) )
net = ll.Conv2DLayer( net, 32, (3,3), nonlinearity=ln.sigmoid )
net = ll.Pool2DLayer( net, pool_size=(2,2) )
net = ll.Conv2DLayer( net, 64, (3,3), nonlinearity=ln.sigmoid )
net = ll.Pool2DLayer( net, pool_size=(2,2) )
net = ll.DenseLayer( net, 64, nonlinearity=ln.sigmoid)
net = ll.DenseLayer( net, 1, nonlinearity=ln.sigmoid)
net = ll.ReshapeLayer( net, (-1,) )
# Predictions:
probs = ll.get_output( net, deterministic=True )
preds = probs > 0.5
#--- Losses --- #
loss = binary_crossentropy(probs, target_var).mean()
acc = T.mean(T.eq(preds, target_var), dtype='float32')
#--- Training settings --- #
params = ll.get_all_params(net, trainable=True)
updates = lasagne.updates.adadelta(loss, params)
#--- Compile Theano functions --- #
train_fn = theano.function([input_var, target_var], [loss, acc], updates=updates)
valid_fn = theano.function([input_var, target_var], [loss, acc])
#--- Training Loop --- #
num_epochs = 100
batch_size = 10
for epoch in range(num_epochs):
# In each epoch, we do a full pass over the training data:
train_err = 0
train_acc = 0
train_batches = 0
start_time = time.time()
for batch in myutils.iterate_minibatches(train_x, train_y, batch_size, shuffle=False):
inputs, targets = batch
err, acc = train_fn(inputs, targets)
train_err += err
train_acc += acc
train_batches += 1
# And a full pass over the validatio data:
val_err = 0
val_acc = 0
val_batches = 0
for batch in myutils.iterate_minibatches(test_x, test_y, batch_size, shuffle=False):
inputs, targets = batch
v_err, v_acc = valid_fn(inputs, targets)
val_err += v_err
val_acc += v_acc
val_batches += 1
# Then we print the results for this epoch:
print("Epoch {} of {} took {:.3f}s".format(
epoch + 1, num_epochs, time.time() - start_time))
print(" training loss:\t\t{:.6f}".format(train_err / train_batches))
print(" training accuracy:\t\t{:.2f} %".format(
train_acc / train_batches * 100))
print(" validation loss:\t\t{:.6f}".format(val_err / val_batches))
print(" validation accuracy:\t\t{:.2f} %".format(
val_acc / val_batches * 100))
# Optionally, you could now dump the network weights to a file like this:
# np.savez('model.npz', *lasagne.layers.get_all_param_values(network))
#
# And load them again later on like this:
# with np.load('model.npz') as f:
# param_values = [f['arr_%d' % i] for i in range(len(f.files))]
# lasagne.layers.set_all_param_values(network, param_values)
#----------------------------------------------------------------------
# Done.
#----------------------------------------------------------------------