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

#*****************************************************************************80
#
## boston_housing_external uses keras to analyze Boston housing data.
#
#  Licensing:
#
#    This code is distributed under the MIT license.
#
#  Modified:
#
#    24 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 numpy as np
  import platform
  
  print ( '' )
  print ( 'boston_housing_external:' )
  print ( '  python version: %s' % ( platform.python_version ( ) ) )
  print ( '  keras version: %s' % ( keras.__version__ ) )
  print ( '  Neural network to classify the Boston housing data.' )
  print ( '  The data is read from external files, not internal datasets.' )
#
#  Load the datasets from external files.
#
#  Here is the statement that would load the data from the built in dataset:
#
# ( train_data, train_targets ), ( test_data, test_targets ) = boston_housing.load_data ( )
#
  data = np.loadtxt ( 'boston_housing.txt' )
  data = data[:,1:]
  print ( data.shape )

  train_index = np.loadtxt ( 'boston_housing_train_index.txt', dtype = 'int' )
  train_index = train_index - 1

  test_index = np.loadtxt ( 'boston_housing_test_index.txt', dtype = 'int' )
  test_index = test_index - 1
#
#  Set up the training and test targets (selling prices)
#
  targets = data[:,13]

  train_targets = targets[train_index]
  print ( train_targets.shape )

  test_targets = targets[test_index]
  print ( test_targets.shape )
#
#  Set up the training and test data.
#
  train_data = data[train_index,:]
  print ( train_data.shape )

  test_data = data[test_index,:]
  print ( test_data.shape )
#
#  The rest of the program is the same as "boston_housing.py"...
#
#  Exhibit the contents of one item of the training data:
#
  print ( train_targets[0:3], train_targets[-3:] )
#
#  Normalize the data.
#
  mean = train_data.mean ( axis = 0 )
  train_data = train_data - mean
  std = train_data.std ( axis = 0 )
  train_data = train_data / std

  test_data = test_data - mean
  test_data = test_data / std
#
#  Define the network architecture.
#
#  Because we are going to instantiate the same model several times, we
#  implement it as a function.
#
  from keras import models
  from keras import layers

  def build_model ( ):
    model = models.Sequential()
    model.add ( layers.Dense ( 64, activation = 'relu', input_shape = ( train_data.shape[1], ) ) )
    model.add ( layers.Dense ( 64, activation = 'relu' ) )
    model.add ( layers.Dense ( 1 ) )
    model.compile ( \
      optimizer = 'rmsprop', \
      loss = 'mse', \
      metrics = [ 'mae' ] )
    return model
#
#  Do K-fold validation
#
  k = 4
  num_val_samples = len ( train_data ) // k
  num_epochs = 100
  all_scores = []

  for i in range ( k ):

    print ( '  Processing fold #', i )

    val_data = train_data[i*num_val_samples:(i+1)*num_val_samples]
    val_targets = train_targets[i*num_val_samples:(i+1)*num_val_samples]

    partial_train_data = np.concatenate ( \
      [ train_data[:i*num_val_samples], \
        train_data[:(i+1)*num_val_samples:]], \
      axis = 0 )
    partial_train_targets = np.concatenate ( \
      [ train_targets[:i*num_val_samples], \
        train_targets[:(i+1)*num_val_samples:]], \
      axis = 0 )

    model = build_model ( )

    model.fit ( \
      partial_train_data, \
      partial_train_targets, \
      epochs = num_epochs, \
      batch_size = 1, \
      verbose = 0 )

    val_mse, val_mae = model.evaluate ( val_data, val_targets, verbose = 0 )

    all_scores.append ( val_mae )

  print ( all_scores )
  print ( np.mean ( all_scores ) )
#
#  Save the validation logs at each fold.
#
  num_epochs = 500
  all_mae_histories = []

  for i in range ( k ):

    print ( '  Processing fold #', i )

    val_data = train_data[i*num_val_samples:(i+1)*num_val_samples]
    val_targets = train_targets[i*num_val_samples:(i+1)*num_val_samples]

    partial_train_data = np.concatenate ( \
      [ train_data[:i*num_val_samples], \
        train_data[:(i+1)*num_val_samples:]], \
      axis = 0 )
    partial_train_targets = np.concatenate ( \
      [ train_targets[:i*num_val_samples], \
        train_targets[:(i+1)*num_val_samples:]], \
      axis = 0 )

    model = build_model ( )

    history = model.fit ( \
      partial_train_data, \
      partial_train_targets, \
      validation_data = ( val_data, val_targets ), \
      epochs = num_epochs, \
      batch_size = 1, \
      verbose = 0 )

    mae_history = history.history [ 'val_mean_absolute_error' ]

    all_mae_histories.append ( mae_history )
#
#  Build the history of successive k-fold validation scores.
#
  average_mae_history = [ \
    np.mean ( [ x[i] for x in all_mae_histories ] ) for i in range ( num_epochs ) ]
#
#  Display the validation scores.
#
  import matplotlib.pyplot as plt

  plt.plot ( range ( 1, len ( average_mae_history ) + 1 ), average_mae_history )
  plt.xlabel ( 'Epochs' )
  plt.ylabel ( 'Validation MAE' )
  plt.title ( 'Evolution of validation MAEs' )
  filename = 'boston_housing_external_validation_mae.png'
  plt.savefig ( filename )
  print ( '' )
  print ( '  Graphics saved in "%s"' % ( filename ) )
  plt.show ( )
#
#  Redisplay, omitting first 10 values.
#
  def smooth_curve ( points, factor = 0.9 ):
    smoothed_points = []
    for point in points:
      if ( smoothed_points ):
        previous = smoothed_points[-1]
        smoothed_points.append ( previous * factor + point * ( 1.0 - factor ) )
      else:
        smoothed_points.append ( point )
    return smoothed_points

  smoothed_mae_history = smooth_curve ( average_mae_history[10:] )

  plt.plot ( range ( 1, len ( smoothed_mae_history ) + 1 ), smoothed_mae_history )
  plt.xlabel ( 'Epochs' )
  plt.ylabel ( 'Validation MAE' )
  plt.title ( 'Evolution of validation MAEs (smoothed)' )
  filename = 'boston_housing_external_validation_mae_smoothed.png'
  plt.savefig ( filename )
  print ( '' )
  print ( '  Graphics saved in "%s"' % ( filename ) )
  plt.show ( )
#
#  Training the final model.
#
  model = build_model ( )
  model.fit ( train_data, train_targets, epochs = 80, batch_size = 16, verbose = 0 )
  test_mse_score, test_mae_score = model.evaluate ( test_data, test_targets )

  print ( test_mae_score )
#
#  Terminate.
#
  print ( '' )
  print ( 'boston_housing_external:' )
  print ( '  Normal end of execution.' )
  return

def timestamp ( ):

#*****************************************************************************80
#
## TIMESTAMP prints the date as a timestamp.
#
#  Licensing:
#
#    This code is distributed under the MIT 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 ( )
  boston_housing_external ( )
  timestamp ( )