Deep Learning and its Applications using Python

2Basics of TensorFlow

This chapter explains TensorFlow fundamentals based on deep learning framework. It plays a major role in pattern recognition, specifically about language, images, sound, and time-series data. Classification, prediction, clustering, and feature extraction have been done with the help of deep learning. Favorably, TensorFlow released in November 2015 by Google and its evolution are tabulated in Table 2.1.

The aim of this chapter is to explain the basic components of TensorFlow. TensorFlow [10] has a facility for performing partial sub-graph computation to agree distributed training by partitioning the neural networks. In addition to that, TensorFlow agrees model parallelism as well as data parallelism. TensorFlow also offers numerous APIs. The deepest level API has named as TensorFlow Core, which provide wide-ranging programming control. The important features of TensorFlow have been enumerated below:

• Its graph deals an illustration of computations.
• Its graph has nodes used for operations.
• It performs computation within stipulated period.
• A graph for computation must be launched in a session.
• The devices such as CPU and GPU places the graph operations in a session.
• For executing graph operations, a session, which has methods have been used.

2.1 Tensors

Initially, the basic of TensorFlow have been discussed. It is a mathematical object. A multidimensional array is used for representation. A tensor [11] of rank one is vector/array whereas tensor of rank ...