Introduction
In the previous chapter, we covered TensorFlow's fundamentals, and we are now able to set up a computational graph. This chapter will introduce Keras, a high-level neural network API written in Python with multiple backends. TensorFlow is one of them. François Chollet, a French software engineer and AI researcher currently working at Google, created Keras for his own personal use before it was open-sourced in 2015. Keras's primary goal is to provide an easy-to-use and accessible library to enable fast experiments.
TensorFlow v1 suffers from usability issues; in particular, a sprawling and sometimes confusing API. For example, TensorFlow v1 offers two high-level APIs:
- The Estimator API (added in release 1.1) is used for training models on localhost or distributed environments
- The Keras API was then added later (release 1.4.0) and intended to be used for fast prototyping
With TensorFlow v2, Keras became the...