-
Notifications
You must be signed in to change notification settings - Fork 6
/
contrasting_cases.py
executable file
·163 lines (133 loc) · 5.29 KB
/
contrasting_cases.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
#!/usr/bin/env python
from __future__ import absolute_import
from __future__ import print_function
import numpy as np
np.random.seed(1337) # for reproducibility
import sys
import argparse
import h5py
from keras.datasets import mnist
from keras.models import Sequential
from keras.layers.core import Dense, Dropout, Activation
from keras.layers.normalization import BatchNormalization
from keras.optimizers import SGD, Adadelta
from keras.utils import np_utils
from outliers import PMeansMultivariateNormal
def create_dataset(n, train_size, valid_size):
means = np.arange(100)
cov = [range(1, 101)] * 100
mvn = PMeansMultivariateNormal(means, cov, (n,))
X = mvn.generate()
assert n % 2 == 0
assert n > train_size + valid_size
# Make the data different along one dimension.
even = np.arange(0, n, step=2)
X[even, 0] = np.random.uniform(-.25, 1.75, size=n/2)
# Make each odd-numbered row the inverse of its previous row.
X[even+1, 0] = np.random.uniform(-1.75, .25, size=n/2)
X += np.random.uniform(0.01, size=X.shape)
X = X.astype(np.float32)
y = np.array([[0,1] * (n/2)]).reshape((n,1))
y = y.astype(np.int32)
X_train = X[0:train_size, :]
X_valid = X[train_size:train_size+valid_size, :]
X_test = X[train_size+valid_size:, :]
y_train = y[0:train_size]
y_valid = y[train_size:train_size+valid_size]
y_test = y[train_size+valid_size:]
return X_train, X_valid, X_test, \
y_train, y_valid, y_test
def build_model(n_inputs, n_hidden, n_classes):
model = Sequential()
model.add(Dense(n_inputs, n_hidden))
model.add(BatchNormalization((n_hidden,)))
model.add(Activation('relu'))
model.add(Dense(n_hidden, n_hidden))
model.add(BatchNormalization((n_hidden,)))
model.add(Activation('relu'))
model.add(Dense(n_hidden, n_hidden))
model.add(BatchNormalization((n_hidden,)))
model.add(Activation('relu'))
model.add(Dense(n_hidden, n_hidden))
model.add(BatchNormalization((n_hidden,)))
model.add(Activation('relu'))
model.add(Dense(n_hidden, n_classes))
model.add(Activation('softmax'))
optimizer = Adadelta()
model.compile(loss='categorical_crossentropy', optimizer=optimizer)
return model
def get_parser():
parser = argparse.ArgumentParser(
description='train a model to demonstrate contrasting cases')
parser.add_argument(
'--shuffle', action='store_true',
help='shuffle the training examples after each epoch (i.e. do not use contrasting cases)')
parser.add_argument(
'--n', type=int, default=10000,
help='the size of the data set to create')
parser.add_argument(
'--train-size', type=int, default=7000,
help='the number of examples from the data set to allocate to training')
parser.add_argument(
'--valid-size', type=int, default=1500,
help='the number of examples from the data set to allocate to validation')
parser.add_argument(
'--batch-size', type=int, default=10,
help='mini-batch size')
parser.add_argument(
'--n-epochs', type=int, default=20,
help='number of epochs to train')
parser.add_argument(
'--verbose', action='store_true',
help='print progress')
return parser.parse_args()
def main(args):
x_train, x_valid, x_test, \
y_train, y_valid, y_test = create_dataset(
args.n, args.train_size, args.valid_size)
y_train = y_train.reshape((y_train.shape[0], 1))
y_valid = y_valid.reshape((y_valid.shape[0], 1))
y_test = y_test.reshape((y_test.shape[0], 1))
n_classes = len(np.unique(y_train))
'''
print('y_train', y_train.shape)
print('y_valid', y_valid.shape)
print('y_test', y_test.shape)
print('n_classes', n_classes, np.unique(y_train))
'''
# convert class vectors to binary class matrices
y_train = np_utils.to_categorical(
y_train, n_classes).astype(np.int32)
y_valid = np_utils.to_categorical(
y_valid, n_classes).astype(np.int32)
y_test = np_utils.to_categorical(
y_test, n_classes).astype(np.int32)
if args.shuffle:
print('Training (shuffled)')
# Leave odd-numbered rows where they are; shuffle only
# even-numbered ones. This ensures that each minibatch has one
# example from each class.
perm = np.arange(x_train.shape[0])
evens = np.arange(0, x_train.shape[0], 2)
perm[evens] = np.random.permutation(evens)
else:
print('Training (contrasting cases)')
model = build_model(100, 20, n_classes)
print('x_train', x_train.dtype)
print('y_train', y_train.dtype)
model.fit(x_train, y_train,
batch_size=args.batch_size,
shuffle=False,
nb_epoch=args.n_epochs,
show_accuracy=True,
verbose=2 if args.verbose else 0,
validation_data=(x_valid, y_valid))
score = model.evaluate(x_test, y_test,
show_accuracy=True,
verbose=1 if args.verbose else 0)
if args.shuffle:
print('Test accuracy (shuffled)', score[1])
else:
print('Test accuracy (contrasting cases)', score[1])
if __name__ == '__main__':
sys.exit(main(get_parser()))