October 2018
Intermediate to advanced
340 pages
7h 56m
English
The TensorFlow library is based on computational graphs, such as a = d * e, d = b + c, and e = c+ 2, and so this formula can be written as a = (b+c) * (c+2), as shown in the following diagram:

The preceding graph computation can be parallelized by executing (d = b + c and e = c + 2) and by splitting the calculations on both CPUs and GPUs. For complex deep learning problems, especially in Convolutional Neural Network (CNNs) and Recurrent Neural Network (RNNs) architectures, this is essential. The concept behind TensorFlow is to have the capability to provision these computational graphs in code and allow ...
Read now
Unlock full access