Additional modules in Keras

Keras provides several additional modules that supplement the basic workflow (described at the beginning of this chapter) with additional functionalities. Some of the modules are as follows:

• The preprocessing module provides several functions for the preprocessing of sequence, image, and text data. 
• The datasets module provides several functions for quick access to several popular datasets, such as CIFAR10 images, CIFAR100 images, IMDB movie reviews, Reuters newswire topics, MNIST handwritten digits, and Boston housing prices.
• The initializers module provides several functions to set initial random weight parameters of layers, such as Zeros, Ones, Constant, RandomNormal, RandomUniform, TruncatedNormal, VarianceScaling, Orthogonal, Identity, lecun_normal, lecun_uniform, glorot_normal, glorot_uniform, he_normal, and he_uniform.
• The models module provides several functions to restore the model architectures and weights, such as model_from_json, model_from_yaml, and load_model...