We have learned that TensorFlow 2.0 uses Keras as a high-level API. Using a high-level API enables rapid prototyping. But we can't use high-level APIs when we want to build a model on a low level, or if we want to build something that a high-level API doesn't provide.

In addition to this, writing code from scratch strengthens our knowledge of algorithms and helps us to understand and learn concepts better than directly diving into high-level APIs. That's why, in this book, we will code most of the algorithms from scratch using TensorFlow, without using high-level APIs such as Keras. We will be using TensorFlow version 1.13.1.