#! /usr/bin/env python3
#
def mnist ( ):

#*****************************************************************************80
#
## mnist uses keras to classify the MNIST digit data.
#
#  Discussion:
#
#    This script is a very simple example that demonstrates the use of keras.
#
#    The MNIST digit dataset, which is built in to keras, is used.
#
#  Licensing:
#
#    This code is distributed under the GNU LGPL license.
#
#  Modified:
#
#    22 April 2019
#
#  Author:
#
#    The original text is by Francois Chollet;
#    some modifications by John Burkardt.
#
#  Reference:
#
#    Francois Chollet,
#    Deep Learning with Python,
#    Manning, 2018,
#    ISBN: 9781617294433.
#
  import keras
  import platform
  
  print ( '' )
  print ( 'mnist:' )
  print ( '  python version: %s' % ( platform.python_version ( ) ) )
  print ( '  keras version: %s' % ( keras.__version__ ) )
  print ( '  Neural network to classify the MNIST digit image data.' )
#
#  Import and load the dataset.
#
  from keras.datasets import mnist

  ( train_images, train_labels ), ( test_images, test_labels ) = mnist.load_data()
#
#  Display the shape, size, and type of the training and testing items.
#
  print ( '' )
  print ( '  Number of training items = %d' % ( len ( train_labels ) ) )
  print ( '  Training data shape = (%d,%d,%d)' % ( train_images.shape ) )
  print ( '  Sample training labels:' )
  print ( train_labels )

  print ( '' )
  print ( '  Number of testing items = %d' % ( len ( test_labels ) ) )
  print ( '  Training data shape = (%d,%d,%d)' % ( train_images.shape ) )
  print ( '  Sample testing labels:' )
  print ( test_labels )
  print ( '' )
#
#  Define the network architecture.
#
  from keras import models
  from keras import layers

  network = models.Sequential()
  network.add ( layers.Dense ( 512, activation = 'relu', input_shape = ( 28 * 28, ) ) )
  network.add ( layers.Dense ( 10, activation = 'softmax' ) )
#
#  Compile the network.
#
  network.compile ( \
    optimizer = 'rmsprop', \
    loss = 'categorical_crossentropy', \
    metrics = [ 'accuracy' ] )
#
#  Convert the training data from uint8 arrays of shape (60000,28,28) to
#  float32 arrays of shape (60000,28*28), and similarly for the test data.
#
  train_images = train_images.reshape ( ( 60000, 28 * 28 ) )
  train_images = train_images.astype ( 'float32' ) / 255
  test_images = test_images.reshape ( ( 10000, 28 * 28 ) )
  test_images = test_images.astype ( 'float32' ) / 255
#
#  Convert the labels from numeric values to categorical values.
#
  from keras.utils import to_categorical

  train_labels = to_categorical ( train_labels )
  test_labels = to_categorical ( test_labels )
#
#  Train the network.
#
  network.fit ( train_images, train_labels, epochs = 5, batch_size = 128 )
#
#  Test the network.
#
  test_loss, test_acc = network.evaluate ( test_images, test_labels )
#
#  Print the test accuracy.
#
  print ( '' )
  print ( 'test_loss: %g' % ( test_loss ) )
  print ( 'test_acc:  %g' % ( test_acc ) )
#
#  Terminate.
#
  print ( '' )
  print ( 'mnist:' )
  print ( '  Normal end of execution.' )
  return

def timestamp ( ):

#*****************************************************************************80
#
## TIMESTAMP prints the date as a timestamp.
#
#  Licensing:
#
#    This code is distributed under the GNU LGPL license. 
#
#  Modified:
#
#    06 April 2013
#
#  Author:
#
#    John Burkardt
#
#  Parameters:
#
#    None
#
  import time

  t = time.time ( )
  print ( time.ctime ( t ) )

  return None

if ( __name__ == '__main__' ):
  timestamp ( )
  mnist ( )
  timestamp ( )