gistfile1.py

from keras.models import *
from keras.optimizers import *
from keras.regularizers import *
from keras.datasets import mnist
from keras.utils import np_utils


def create_model():
    x = Input(shape=(784,), name='x')

    h = BatchNormalization(mode=2)(x)
    y = Dense(64, init='uniform', activation='relu', W_regularizer='l2', b_regularizer='l2')(h)

    model = Model(input=x, output=y)

    x1 = Input(shape=(784,), name='x1')
    x2 = Input(shape=(784,), name='x2')
    y1 = model(x1)
    y2 = model(x2)

    merged_vector = merge([y1, y2], mode=lambda x: x[0] - x[1], output_shape=(64,))
    output = Dense(10, activation='sigmoid')(merged_vector)

    net = Model(input=[x1, x2], output=output)
    net.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
    return model, net

batch_size = 128
nb_classes = 10
nb_epoch = 1

# the data, shuffled and split between train and test sets
(X_train, y_train), (X_test, y_test) = mnist.load_data()

X_train = X_train.reshape(60000, 784)
X_test = X_test.reshape(10000, 784)
X_train = X_train.astype('float32')
X_test = X_test.astype('float32')
X_train /= 255
X_test /= 255
print(X_train.shape[0], 'train samples')
print(X_test.shape[0], 'test samples')

# convert class vectors to binary class matrices
Y_train = np_utils.to_categorical(y_train, nb_classes)
Y_test = np_utils.to_categorical(y_test, nb_classes)

model, net = create_model()
json_string = net.to_json()
open('my_model_architecture.json', 'w').write(json_string)

history = net.fit([X_train, X_train], Y_train,
                    batch_size=batch_size, nb_epoch=nb_epoch,
                    verbose=1, validation_data=([X_test, X_test], Y_test))