#!/usr/bin/env python
# coding: utf-8

# In[15]:


from keras.datasets import mnist

# use Keras to import pre-shuffled MNIST database
(X_train, y_train), (X_test, y_test) = mnist.load_data()

print("The MNIST database has a training set of %d examples." % len(X_train))
print("The MNIST database has a test set of %d examples." % len(X_test))


# In[16]:


import matplotlib.pyplot as plt
get_ipython().run_line_magic('matplotlib', 'inline')
import matplotlib.cm as cm
import numpy as np

# plot first six training images
fig = plt.figure(figsize=(20,20))
for i in range(6):
    ax = fig.add_subplot(1, 6, i+1, xticks=[], yticks=[])
    ax.imshow(X_train[i], cmap='gray')
    ax.set_title(str(y_train[i]))


# In[17]:


def visualize_input(img, ax):
    ax.imshow(img, cmap='gray')
    width, height = img.shape
    thresh = img.max()/2.5
    for x in range(width):
        for y in range(height):
            ax.annotate(str(round(img[x][y],2)), xy=(y,x),
                        horizontalalignment='center',
                        verticalalignment='center',
                        color='white' if img[x][y]<thresh else 'black')

fig = plt.figure(figsize = (12,12)) 
ax = fig.add_subplot(111)
visualize_input(X_train[0], ax)


# In[18]:


# rescale [0,255] --> [0,1]
X_train = X_train.astype('float32')/255
X_test = X_test.astype('float32')/255


# In[19]:


from keras.utils import np_utils

# print first ten (integer-valued) training labels
print('Integer-valued labels:')
print(y_train[:10])

# one-hot encode the labels
y_train = np_utils.to_categorical(y_train, 10)
y_test = np_utils.to_categorical(y_test, 10)

# print first ten (one-hot) training labels
print('One-hot labels:')
print(y_train[:10])


# In[20]:


from keras.models import Sequential
from keras.layers import Dense, Dropout, Flatten

# define the model
model = Sequential()
model.add(Flatten(input_shape=X_train.shape[1:]))
model.add(Dense(512, activation='relu'))
model.add(Dropout(0.2))
model.add(Dense(512, activation='relu'))
model.add(Dropout(0.2))
model.add(Dense(10, activation='softmax'))

# summarize the model
model.summary()


# In[21]:


# compile the model
model.compile(loss='categorical_crossentropy', optimizer='rmsprop', 
              metrics=['accuracy'])


# In[22]:


# evaluate test accuracy
score = model.evaluate(X_test, y_test, verbose=0)
accuracy = 100*score[1]

# print test accuracy
print('Test accuracy: %.4f%%' % accuracy)


# In[23]:


from keras.callbacks import ModelCheckpoint   

# train the model
checkpointer = ModelCheckpoint(filepath='mnist.model.best.hdf5', 
                               verbose=1, save_best_only=True)
hist = model.fit(X_train, y_train, batch_size=128, epochs=10,
          validation_split=0.2, callbacks=[checkpointer],
          verbose=1, shuffle=True)


# In[24]:


# load the weights that yielded the best validation accuracy
model.load_weights('mnist.model.best.hdf5')


# In[25]:


# evaluate test accuracy
score = model.evaluate(X_test, y_test, verbose=0)
accuracy = 100*score[1]

# print test accuracy
print('Test accuracy: %.4f%%' % accuracy)