-
Notifications
You must be signed in to change notification settings - Fork 0
/
nnet.py
134 lines (108 loc) · 5.03 KB
/
nnet.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
# ref : https://github.com/ottogroup/kaggle/blob/master/Otto_Group_Competition.ipynb
# ref : https://gist.github.com/dnouri/fe855653e9757e1ce8c4
import common
import smote
import numpy as np
from lasagne import layers
from lasagne.layers import DenseLayer, InputLayer, DropoutLayer
from lasagne.nonlinearities import rectify, softmax, tanh, linear
from lasagne.updates import nesterov_momentum, rmsprop, momentum
from lasagne.objectives import categorical_crossentropy
from nolearn.lasagne import NeuralNet
from sklearn.grid_search import GridSearchCV, RandomizedSearchCV
from scipy.stats import uniform as sp_rand
from scipy.stats import randint as sp_randint
import matplotlib
import matplotlib.pyplot as plt
import matplotlib.cm as cm
from sklearn.metrics import classification_report
from sklearn.metrics import confusion_matrix
# ref : https://github.com/dnouri/nolearn/issues/95
def lasagne_grid_search(X, y):
param_dist = {"update_learning_rate": [0.001, 0.01],
"dense0_num_units": [100, 300],
"dense1_num_units": [100, 300],
"dropout0_p": [0.1, 0.75],
"dropout1_p": [0.1, 0.75],
"update_momentum" : [0.9, 0.99] #,
}
# run 1 best : [CV] update_learning_rate=0.001, dropout1_p=0.1, update_momentum=0.99, dropout0_p=0.1, dense1_num_units=300, dense0_num_units=300, score=0.603377 - 16.8s
param_dist = {"update_learning_rate": [0.0001, 0.001],
"dense0_num_units": [300, 600],
"dense1_num_units": [300, 600],
"dropout0_p": [0.05, 0.1],
"dropout1_p": [0.05, 0.1],
"update_momentum" : [0.99, 0.9999]
}
# run 2 best : [CV] update_learning_rate=0.001, dropout1_p=0.1, update_momentum=0.99, dropout0_p=0.05, dense1_num_units=600, dense0_num_units=300, score=0.605281 - 25.1s
param_dist = {"update_learning_rate": [0.001],
"dense0_num_units": [300],
"dense1_num_units": [450, 600, 750],
"dropout0_p": [0.025, 0.05, 0.075, 0.1, 0.25],
"dropout1_p": [0.1],
"update_momentum" : [0.99]
}
# run 3 best : [CV] update_learning_rate=0.001, dropout1_p=0.1, update_momentum=0.99, dropout0_p=0.075, dense1_num_units=750, dense0_num_units=300, score=0.605437 - 29.5s
layers0 = [('input', InputLayer),
('dense0', DenseLayer),
('dropout0', DropoutLayer),
('dense1', DenseLayer),
('dropout1', DropoutLayer),
('output', DenseLayer)]
net0 = NeuralNet(layers=layers0,
input_shape=(None, X.shape[1]),
output_num_units=3,
output_nonlinearity=softmax,
## objective_loss_function =categorical_crossentropy
update=nesterov_momentum,
eval_size=0.2,
verbose=1,
max_epochs=20
)
random_search = GridSearchCV(net0, param_grid=param_dist, cv=2, verbose=4) #, scoring=NeuralNet.score , cv=2, n_jobs=1, verbose=5, refit=False)
random_search.fit(X, y)
def classify(X, y, X_test, y_test):
layers0 = [('input', InputLayer),
('dense0', DenseLayer),
('dropout0', DropoutLayer),
('dense1', DenseLayer),
('dropout1', DropoutLayer),
('output', DenseLayer)]
net = NeuralNet(layers=layers0,
input_shape=(None, X.shape[1]),
dense0_num_units=300,
dropout0_p=0.075,
dropout1_p=0.1,
dense1_num_units=750,
output_num_units=3,
output_nonlinearity=softmax,
update=nesterov_momentum,
update_learning_rate=0.001,
update_momentum=0.99,
eval_size=0.2,
verbose=1,
max_epochs=15)
net.fit(X, y)
print(net.score(X, y))
preds = net.predict(X_test)
print(classification_report(y_test, preds))
cm = confusion_matrix(y_test, preds)
plt.matshow(cm)
plt.title('Confusion matrix')
plt.colorbar()
plt.ylabel('True label')
plt.xlabel('Predicted label')
plt.savefig('confmatrix.png')
plt.show()
print(cm)
def main():
data_train, data_test, target_train, target_test = common.load_train_data_and_split(file='data/processed_missing_filled_in.csv')
data_train = np.asarray(data_train)
target_train = np.array(target_train)
target_train = target_train.astype(np.int32)
print(target_train)
data_train, target_train = smote.smote_data(data_train, target_train)
classify(data_train, target_train, data_test, target_test)
# lasagne_grid_search(data_train, target_train)
if __name__ == '__main__':
main()