MNIST Hierarchical RNN



In [ ]:

    
# from: https://github.com/fchollet/keras/blob/master/examples/mnist_hierarchical_rnn.py



In [1]:

    
from __future__ import print_function

from keras.datasets import mnist
from keras.models import Sequential, Model
from keras.layers import Input, Dense, TimeDistributed
from keras.layers import LSTM
from keras.utils import np_utils









    



Using TensorFlow backend.



In [2]:

    
# Training parameters.
batch_size = 32
nb_classes = 10
nb_epochs = 5



In [3]:

    
# Embedding dimensions.
row_hidden = 128
col_hidden = 128



In [4]:

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



In [5]:

    
X_train?



In [6]:

    
# Reshapes data to 4D for Hierarchical RNN.
X_train = X_train.reshape(X_train.shape[0], 28, 28, 1)
X_test = X_test.reshape(X_test.shape[0], 28, 28, 1)
X_train = X_train.astype('float32')
X_test = X_test.astype('float32')
X_train /= 255
X_test /= 255
print('X_train shape:', X_train.shape)
print(X_train.shape[0], 'train samples')
print(X_test.shape[0], 'test samples')









    



X_train shape: (60000, 28, 28, 1)
60000 train samples
10000 test samples



In [7]:

    
# Converts 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)



In [8]:

    
row, col, pixel = X_train.shape[1:]



In [9]:

    
# 4D input.
x = Input(shape=(row, col, pixel))



In [10]:

    
# Encodes a row of pixels using TimeDistributed Wrapper.
encoded_rows = TimeDistributed(LSTM(output_dim=row_hidden))(x)



In [11]:

    
# Encodes columns of encoded rows.
encoded_columns = LSTM(col_hidden)(encoded_rows)



In [12]:

    
# Final predictions and model.
prediction = Dense(nb_classes, activation='softmax')(encoded_columns)
model = Model(input=x, output=prediction)
model.compile(loss='categorical_crossentropy',
              optimizer='rmsprop',
              metrics=['accuracy'])



In [13]:

    
# Training.
model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epochs,
          verbose=1, validation_data=(X_test, Y_test))









    



Train on 60000 samples, validate on 10000 samples
Epoch 1/5
60000/60000 [==============================] - 861s - loss: 0.2807 - acc: 0.9097 - val_loss: 0.1012 - val_acc: 0.9713
Epoch 2/5
60000/60000 [==============================] - 876s - loss: 0.0976 - acc: 0.9712 - val_loss: 0.0769 - val_acc: 0.9771
Epoch 3/5
60000/60000 [==============================] - 769s - loss: 0.0707 - acc: 0.9786 - val_loss: 0.0578 - val_acc: 0.9831
Epoch 4/5
60000/60000 [==============================] - 755s - loss: 0.0582 - acc: 0.9831 - val_loss: 0.0797 - val_acc: 0.9751
Epoch 5/5
60000/60000 [==============================] - 782s - loss: 0.0492 - acc: 0.9854 - val_loss: 0.0478 - val_acc: 0.9863






    Out[13]:





<keras.callbacks.History at 0x7fd04c1e3ad0>



In [ ]: