Let's use the autoencoder with MNIST data: mnist = input_data.read_data_sets('MNIST_data', one_hot = True).
Use StandardScalar from Scikit Learn's sklearn.preprocessing module to extract testmnist.test.images and training images mnist.train.images:
X_train, X_test = self.standard_scale(mnist.train.images, mnist.test.images).
The preprocessing module provides a utility class, StandardScaler, which implements the Transformer API. This computes and transforms the mean and standard deviation of a training set. It reapplies the same transformation to the testing set. By default, Scalar centers the mean and makes the variance one.
It is possible to disable either centering or scaling by passing with_mean=False ...
