def create_columns(self, columns=5):
for w in self.Ws:
x_train = self.train_datasets[w]
self.dnns[w] = []
for i in range(columns):
model = DNN(width=x_train.shape[1],
height=x_train.shape[2],
depth=x_train.shape[3],
classes=10)
model.compile(loss="categorical_crossentropy",
optimizer=Adadelta(),
metrics=["accuracy"])
self.dnns[w].append(model)
import numpy as np
from dnn import DNN
from keras.datasets import mnist
from keras.utils import np_utils, to_categorical
(x_train, y_train), (x_test, y_test) = mnist.load_data()
mnist_original_size = 28
x_train = x_train.reshape(x_train.shape[0], mnist_original_size, mnist_original_size, 1)
x_test = x_test.reshape(x_test.shape[0], mnist_original_size, mnist_original_size, 1)
y_train = to_categorical(y_train, 10)
y_test = to_categorical(y_test, 10)
model1 = DNN(width=x_train.shape[1], height=x_train.shape[2], depth=x_train.shape[3], classes=10)
model1.compile(loss="categorical_crossentropy", optimizer='adadelta', metrics=["accuracy"])
model1.fit(x_train, y_train, epochs=1, verbose=1)
loss, accuracy = model1.evaluate(x_test, y_test, verbose=1)
print('\nloss: {:.2f}%, accuracy: {:.2f}%'.format(loss*100, accuracy*100))
m1 = model1.predict_proba(x_test)
model2 = DNN(width=x_train.shape[1], height=x_train.shape[2], depth=x_train.shape[3], classes=10)
model2.compile(loss="categorical_crossentropy", optimizer='adadelta', metrics=["accuracy"])
model2.fit(x_train, y_train, nb_epoch=1, verbose=1)
loss, accuracy = model2.evaluate(x_test, y_test, verbose=1)
print('\nloss: {:.2f}%, accuracy: {:.2f}%'.format(loss*100, accuracy*100))
