from keras import models
from keras import layers
from keras.callbacks import EarlyStopping, ModelCheckpoint
from keras.applications import MobileNet
from keras.models import Model
from keras.applications.vgg16 import VGG16
from keras.layers import BatchNormalization, Flatten, Dense, LeakyReLU, Dropout, GaussianNoise
from tensorflow import set_random_seed
from keras.optimizers import Adam, RMSprop
from keras_gradient_noise import add_gradient_noise
from keras.initializers import glorot_uniform, glorot_normal, he_uniform, he_normal
set_random_seed(1724)
model = models.Sequential()
model.add(
Dense(128, kernel_initializer="glorot_normal", input_shape=(9 * 128, )))
model.add(LeakyReLU())
model.add(Dropout(.15))
model.add(BatchNormalization())
# model.add(GaussianNoise(1))
model.add(Dense(64))
model.add(LeakyReLU())
model.add(BatchNormalization())
model.add(Dense(6, activation='softmax'))
# In[85]:
# Let's use a different optimizer this time
noisy = add_gradient_noise(RMSprop)
model.compile(
optimizer="Adamax",
def perform_learning(train_features,
test_features,
f_output,
local,
gglobal,
deep=False):
print(train_features[0])
if (local and gglobal):
train = [x[1:-1] for x in train_features]
test = [x[1:-1] for x in test_features]
elif (local and not gglobal):
train = [x[1:5] for x in train_features]
test = [x[1:5] for x in test_features]
elif (not local and gglobal):
train = [x[5:-1] for x in train_features]
test = [x[5:-1] for x in test_features]
# print train[0]
train_tags = [x[-1] for x in train_features]
test_tags = [x[-1] for x in test_features]
train = z_scoring(train)
test = z_scoring(test)
print(len(train[0]))
print(train[0])
if not deep:
algos = ['adaBoost', 'RF', 'L-SVM', 'RBF-SVM', 'SGD']
# algos = ['RBF-SVM']
for algo in algos:
print(algo)
f_output.writelines(algo + '\n')
if algo == 'adaBoost':
clf = AdaBoostClassifier(n_estimators=100)
if algo == 'RF':
clf = RandomForestClassifier(n_estimators=1000,
criterion="gini",
min_samples_split=15,
oob_score=True,
class_weight='balanced',
max_depth=3)
if algo == 'L-SVM':
clf = SVC(kernel='linear',
class_weight="balanced",
C=0.01,
probability=True)
if algo == 'RBF-SVM':
clf = SVC(class_weight="balanced", C=0.01, probability=True)
if algo == 'SGD':
clf = SGDClassifier(alpha=0.0001,
average=False,
class_weight=None,
epsilon=0.1,
eta0=0.0,
fit_intercept=True,
l1_ratio=0.15,
learning_rate='optimal',
loss='hinge',
n_iter=5,
n_jobs=1,
penalty='l2',
power_t=0.5,
random_state=None,
shuffle=True,
verbose=0,
warm_start=False)
# print train
clf.fit(train, train_tags)
if (algo == 'RF'):
print(len(clf.feature_importances_))
print(clf.feature_importances_)
f_output.writelines(str(clf.feature_importances_) + '\n')
evaluate_auc(clf, test, test_tags, train, train_tags, f_output)
else:
print(train[0])
from keras.models import Sequential
from keras.layers import Dense, Dropout
from keras.regularizers import l2, l1_l2
clf = Sequential()
clf.add(
Dense(100,
activation="relu",
kernel_initializer="he_normal",
input_dim=train.shape[1]))
# self.classifier.add(Dropout(0.5))
# self.classifier.add(Dense(100, init='he_normal', activation='relu', W_regularizer=l2(0.5)))
clf.add(Dropout(0.1))
clf.add(
Dense(1,
init='uniform',
activation='sigmoid',
W_regularizer=l1_l2(0.2)))
clf.compile(loss='binary_crossentropy',
optimizer='adam',
metrics=['accuracy'])
clf.fit(train,
train_tags,
validation_data=[test, test_tags],
epochs=100,
batch_size=10,
verbose=2)
evaluate_auc(clf, test, test_tags, train, train_tags, f_output)
