Tensorflowのexamples.tutorials.mnistにあるinput_dataデータクラスの簡単な分析

tensorflowの冒頭で使うデータセットは最も古典的なmnistの手書き文字認識データセットで、tensorflowに直接カプセル化されているのはmnist handwritten character datasetクラスなので、input_dataのメソッドで直接呼び出せるのは便利です。 * データを読み込む、データラベルを読み込む、next_batch()でデータフィードするなどの操作はできますが、後でさらに深化させる場合、カプセル化されたデータクラスのメソッドを使わずに独自のデータセットを学習・テストしたい場合は、この時点で独自のデータセットを作成する必要があります。そこで、まずは公式の tensorflow.examples.tutorials.mnist import input_data インスピレーションを見つける

まず、tensorflowの公式GitHubで、そのimportパッケージのパスに従って、その定義を以下のサイトで見つけます。

次にコード定義を探します。

次に、mnist.pyのコードを開いて、データセットのクラスとメソッドを定義します。

そのDataSet(object)クラスとメソッドは、以下のように定義されていることがわかります。

1. DataSet(object)クラスを定義する。

class DataSet(object):
  """Container class for a dataset (deprecated).
  This CLASS IS DEPRECATED. see
  [contrib/learn/README.md](https://www.tensorflow.org/code/tensorflow/contrib/learn/README.md)
  for general migration instructions.
  """
  def __init__(self,
               images,
               labels,
               fake_data=False,
               one_hot=False,
               dtype=dtypes.float32,
               reshape=True,
               seed=None):
    """Construct a DataSet class.
      Use the one_hot (unique hot encoding) parameter only if fake_data is true.
     `dtype` can be `uint8`, which keeps the input as `[0,255]`
        It can also be `float32` which rescales to `[0,1]`. 
        The seedseed parameter provides a convenient deterministic test.
    """
    seed1, seed2 = random_seed.get_seed(seed)
    # If op level seed is not set, use whatever graph level seed is returned
    numpy.random.seed(seed1 if seed is None else seed2)
    dtype = dtypes.as_dtype(dtype).base_dtype
    if dtype not in (dtypes.uint8, dtypes.float32):
      raise TypeError(
          'Invalid image dtype %r, expected uint8 or float32' % dtype)
    if fake_data:
      self._num_examples = 10000
      self.one_hot = one_hot
    else:
      assert images.shape[0] == labels.shape[0], (
          'images.shape: %s labels.shape: %s' % (images.shape, labels.shape))
      self._num_examples = images.shape[0]

      # Convert the input type [num examples, rows, columns, depth]
      # Convert to [num examples, rows*columns] (assuming depth == 1)
      if reshape:
        assert images.shape[3] == 1
        images = images.reshape(images.shape[0],
                                images.shape[1] * images.shape[2])
      if dtype == dtypes.float32:
        # convert data type [0, 255] -> scale [0.0, 1.0].
        images = images.astype(numpy.float32)
        images = numpy.multiply(images, 1.0 / 255.0)
    self._images = images
    self._labels = labels
    self._epochs_completed = 0
    self._index_in_epoch = 0

  @property
  def images(self):
    return self._images

  @property
  def labels(self):
    return self._labels

  @property
  def num_examples(self):
    return self._num_examples

  @property
  def epochs_completed(self):
    return self._epochs_completed

2. next_batchメソッドを定義します。

 ポイントは、この時点でデータセットがどこから読み込まれたかを記録しておくこと、そして、反復処理ごとにデータセットが（index_in_epoch+batch_size）であることです。そして、最初のエポックをどうするか、各エポックの終わりが次のエポックの始まりに合流するのをどうするか、最初のエポックと終わり以外をどうするか、です。 

  def next_batch(self, batch_size, fake_data=False, shuffle=True):
    """Returns the next `batch_size` data from this dataset, based on the current dataset. """
    if fake_data:
      fake_image = [1] * 784
      if self.one_hot:
        fake_label = [1] + [0] * 9
      else:
        fake_label = 0
      return [fake_image for _ in xrange(batch_size)], [
          fake_label for _ in xrange(batch_size)
      start]
    start = self._index_in_epoch
    # Process epoch start position
    # Shuffle at the beginning to disrupt the data order 
    if self._epochs_completed == 0 and start == 0 and shuffle:
      perm0 = numpy.range(self._num_examples)
      numpy.random.shuffle(perm0)
      self._images = self.images[perm0]
      self._labels = self.labels[perm0]
    # Execute the next iteration cycle
    if start + batch_size > self._num_examples:
      # Complete an iteration cycle
      self._epochs_completed += 1
      # Get the rest examples in this epoch
      rest_num_examples = self._num_examples - start
      images_rest_part = self._images[start:self._num_examples]
      labels_rest_part = self._labels[start:self._num_examples]
      # Shuffle is used to break up the order of the data
      if shuffle:
        perm = numpy.range(self._num_examples)
        numpy.random.shuffle(perm)
        self._images = self.images[perm]
        self._labels = self.labels[perm]
      # Start the next iteration epoch
      start = 0
      self._index_in_epoch = batch_size - rest_num_examples
      end = self._index_in_epoch
      images_new_part = self._images[start:end]
      labels_new_part = self._labels[start:end]
      return numpy.concatenate(
          (images_rest_part, images_new_part), axis=0), numpy.concatenate(
              (labels_rest_part, labels_new_part), axis=0)

   # except for the first epoch, and the beginning of each epoch, the rest of the middle batch batches are processed
     else:
      self._index_in_epoch += batch_size# starting position: index_in_epoch
      end = self._index_in_epoch# end position: start position (index_in_epoch) plus batch_size
      return self._images[start:end], self._labels[start:end]# return images and their corresponding labels

3. read_data_set メソッドを定義する。

def read_data_sets(train_dir,
                   fake_data=False,
                   one_hot=False,
                   dtype=dtypes.float32,
                   reshape=True,
                   validation_size=5000,
                   seed=None,
                   source_url=DEFAULT_SOURCE_URL):
  if fake_data:

    def fake():
      return DataSet(
          [], [], fake_data=True, one_hot=one_hot, dtype=dtype, seed=seed)

    train = fake()
    validation = fake()
    test = fake()
    # Return the data class.
    return base.Datasets(train=train, validation=validation, test=test)

  if not source_url: # empty string check
    source_url = DEFAULT_SOURCE_URL

  TRAIN_IMAGES = 'train-images-idx3-ubyte.gz'
  TRAIN_LABELS = 'train-labels-idx1-ubyte.gz'
  TEST_IMAGES = 't10k-images-idx3-ubyte.gz'
  TEST_LABELS = 't10k-labels-idx1-ubyte.gz'
  #base module is defined in: from tensorflow.contrib.learn.python.learn.datasets import base
  local_file = base.maybe_download(TRAIN_IMAGES, train_dir,
                                   source_url + TRAIN_IMAGES)
  #gfile module is defined in: from tensorflow.python.platform import gfile
  with gfile.Open(local_file, 'rb') as f:
    train_images = extract_images(f)

  local_file = base.be_download(TRAIN_LABELS, train_dir,
                                   source_url + TRAIN_LABELS)
  with gfile.Open(local_file, 'rb') as f:
    train_labels = extract_labels(f, one_hot=one_hot)

  local_file = base.maybe_download(TEST_IMAGES, train_dir,
                                   source_url + TEST_IMAGES)
  with gfile.Open(local_file, 'rb') as f:
    test_images = extract_images(f)

  local_file = base.maybe_download(TEST_LABELS, train_dir,
                                   source_url + TEST_LABELS)
  with gfile.Open(local_file, 'rb') as f:
    test_labels = extract_labels(f, one_hot=one_hot)

  if not 0 <= validation_size <= len(train_images):
    raise ValueError('Validation size should be between 0 and {}. Received: {}.'
                     .format(len(train_images), validation_size))

  validation_images = train_images[:validation_size]
  validation_labels = train_labels[:validation_size]
  train_images = train_images[validation_size:]
  train_labels = train_labels[validation_size:]

  options = dict(dtype=dtype, reshape=reshape, seed=seed)

 # After reading the data return the data classes for the training, validation and test sets (with their methods and properties) according to the DataSet class defined above.

  train = DataSet(train_images, train_labels, **options)
  validation = DataSet(validation_images, validation_labels, **options)
  test = DataSet(test_images, test_labels, **options)

  return base.Datasets(train=train, validation=validation, test=test)

def load_mnist(train_dir='MNIST-data'):
  #return the load data function
  return read_data_sets(train_dir)

参考ブログ https://blog.csdn.net/u013608336/article/details/78747102