Question

我有一个回归任务数据。独立功能（X_train）用标准缩放器缩放。建立了Keras顺序模型，添加了隐藏层。编译模型。然后使用model.fit(X_train_scaled, y_train )拟合模型然后，将模型保存到.hdf5文件中。

现在如何将缩放部分包括在已保存的模型中，以便将相同的缩放参数应用于看不见的测试数据。

#imported all the libraries for training and evaluating the model
X_train, X_test, y_train, y_test = train_test_split( X, y, test_size=0.3, random_state=42)
sc = StandardScaler()
X_train_scaled = sc.fit_transform(X_train)
X_test_scaled= sc.transform (X_test)



def build_model():
    model = keras.Sequential([layers.Dense(64, activation=tf.nn.relu,input_shape=[len(train_dataset.keys())]),
    layers.Dense(64, activation=tf.nn.relu),
    layers.Dense(1)
    ])

    optimizer = tf.keras.optimizers.RMSprop(0.001)

    model.compile(loss='mean_squared_error',
                optimizer=optimizer,
                metrics=['mean_absolute_error', 'mean_squared_error'])
    return model
model = build_model()
EPOCHS=1000
history = model.fit(X_train_scaled, y_train, epochs=EPOCHS,
                    validation_split = 0.2, verbose=0)

loss, mae, mse = model.evaluate(X_test_scaled, y_test, verbose=0)

Answer 1

据我所知，标准和有效的方法是使用Tensorflow Transform。这并不是说如果必须使用TF Transform，就应该使用整个TFX Pipeline。 TF转换也可以用作独立版本。

Tensorflow变换创建一个波束变换图，该图将这些变换作为常量注入Tensorflow图中。由于这些转换在图表中表示为常量，因此它们在训练和服务过程中将保持一致。在培训和服务过程中保持一致性的优势是

消除了培训服务中的偏差
消除了在服务系统中添加代码的需求，从而缩短了等待时间。

TF转换的示例代码如下：

用于导入所有依赖项的代码：

try:
  import tensorflow_transform as tft
  import apache_beam as beam
except ImportError:
  print('Installing TensorFlow Transform.  This will take a minute, ignore the warnings')
  !pip install -q tensorflow_transform
  print('Installing Apache Beam.  This will take a minute, ignore the warnings')
  !pip install -q apache_beam
  import tensorflow_transform as tft
  import apache_beam as beam

import tensorflow as tf
import tensorflow_transform.beam as tft_beam
from tensorflow_transform.tf_metadata import dataset_metadata
from tensorflow_transform.tf_metadata import dataset_schema

下面提到的是预处理功能，其中我们提到了所有转换：

def preprocessing_fn(inputs):
  """Preprocess input columns into transformed columns."""
  # Since we are modifying some features and leaving others unchanged, we
  # start by setting `outputs` to a copy of `inputs.
  outputs = inputs.copy()

  # Scale numeric columns to have range [0, 1].
  for key in NUMERIC_FEATURE_KEYS:
    outputs[key] = tft.scale_to_0_1(outputs[key])

  for key in OPTIONAL_NUMERIC_FEATURE_KEYS:
    # This is a SparseTensor because it is optional. Here we fill in a default
    # value when it is missing.
    dense = tf.sparse_to_dense(outputs[key].indices,
                               [outputs[key].dense_shape[0], 1],
                               outputs[key].values, default_value=0.)
    # Reshaping from a batch of vectors of size 1 to a batch to scalars.
    dense = tf.squeeze(dense, axis=1)
    outputs[key] = tft.scale_to_0_1(dense)

  return outputs

除了

tft.scale_to_0_1

您还可以使用其他API进行标准化，例如

tft.scale_by_min_max, tft.scale_to_z_score

您可以参考下面提到的链接以获取详细信息和TF转换教程。

https://www.tensorflow.org/tfx/transform/get_started

https://www.tensorflow.org/tfx/tutorials/transform/census

如何在Keras回归模型中包括特征的归一化？

1 个答案: