def createModel(dense1, dropout1, dense2, dropout2):
model = Sequential()
model.add(Dense(dense1, input_shape=(dims,), init='he_uniform', W_regularizer=regularizers.l1(0.0005)))
model.add(Activation('relu'))
model.add(Dropout(dropout1))# input dropout
model.add(Dense(dense2, init='he_uniform'))
model.add(Activation('relu'))
model.add(Dropout(dropout2))
model.add(Dense(2, init='he_uniform'))
model.add(Activation('sigmoid'))
model.compile(loss='binary_crossentropy', optimizer="adagrad")
return model
def createModel(x):
model = Sequential()
model.add(Dense(x[0], input_shape=(dims,), init='he_uniform', W_regularizer=regularizers.l1(0.0005)))
model.add(Activation('relu'))
model.add(Dropout(x[1]))# input dropout
model.add(Dense(x[2], init='he_uniform'))
model.add(Activation('relu'))
model.add(Dropout(x[3]))
model.add(Dense(nb_classes, init='he_uniform'))
model.add(Activation('sigmoid'))
model.compile(loss='binary_crossentropy', optimizer="adagrad")
return model
def createModel():
model = Sequential()
model.add(Dense(int(85), input_shape=(dims,), init='he_uniform', W_regularizer=regularizers.l1(0.0005)))
model.add(Dropout(0.05)) # input dropout
model.add(PReLU())
model.add(BatchNormalization())
model.add(Dropout(0.1))
model.add(Dense(int(83)))
model.add(PReLU())
model.add(BatchNormalization())
model.add(Dropout(0.15))
model.add(Dense(nb_classes))
model.add(Activation('sigmoid'))
model.compile(loss='binary_crossentropy', optimizer="adam")
return model
def createModel(dense1, dropout1, dense2, dropout2):
model = Sequential()
model.add(
Dense(dense1,
input_shape=(dims, ),
init='he_uniform',
W_regularizer=regularizers.l1(0.0005)))
model.add(Activation('relu'))
model.add(Dropout(dropout1)) # input dropout
model.add(Dense(dense2, init='he_uniform'))
model.add(Activation('relu'))
model.add(Dropout(dropout2))
model.add(Dense(2, init='he_uniform'))
model.add(Activation('sigmoid'))
model.compile(loss='binary_crossentropy', optimizer="adagrad")
return model
def kerascv(dense1, dense2, epochs):
ival = IntervalEvaluation(validation_data=(x1, y1), interval=1)
pred_sum = 0
for k in range(1):
model = Sequential()
model.add(
Dense(int(dense1),
input_shape=(dims, ),
init='he_uniform',
W_regularizer=regularizers.l1(0.0005)))
model.add(Dropout(0.05)) # input dropout
model.add(PReLU())
model.add(BatchNormalization())
model.add(Dropout(0.1))
model.add(Dense(int(dense2)))
model.add(PReLU())
model.add(BatchNormalization())
model.add(Dropout(0.15))
model.add(Dense(nb_classes))
model.add(Activation('sigmoid'))
model.compile(loss='binary_crossentropy', optimizer="adam")
model.fit(x0,
y0,
nb_epoch=int(epochs),
batch_size=128,
verbose=0,
callbacks=[ival])
preds = model.predict_proba(x1, batch_size=64, verbose=0)[:, 1]
pred_sum += preds
pred_average = pred_sum / (k + 1)
del model
loss = auc(y1[:, 1], pred_average)
return loss