In TensorFlow, computational operations are interdependent and thus interact with one another. Computational graphs help to track these interdependencies, thus making sense of a complicated functional architecture.
Computation graphs
What is a computation graph?
Each node in the TensorFlow graph is a symbolic representation of an operational process. When the data reaches a particular node during the flow, the corresponding operational function associated with the node is performed. The output of the process is fed to the next node as an input.
The primary benefit of graph computation is that it helps to perform optimized computations:
