I will walk you through the basics with 3 question answers. But Before we start learning about tensorflow, first of all I need you to answer few questions.
Who created this API ?
-> It is built by the Google Brain team for solving their Deep Learning problems.
Is it just a mere API ?
-> It is actually much more than that
What special it does ?
-> It creates just a graph until the code is run using sessions. So on every run the data flows through this graphical pipeline and gives results
So now that you have a fair bit of Idea what exactly this does, so we can move further to our n number of question answers. We will be using python for playing with TensorFlow in this article.
How to install Install Tensorflow on linux, Mac and Windows?
Two versions of tensorflow are available :
- CPU version :- pip install –upgrade tensorfow
- GPU version :- pip install –upgrade tensorfow-gpu
What is ABC of TensorFlow?
- Variable and Placeholder : Variables can be understood as the usual set of variables which we use in all the programming languages whose value can be changed later during the program but they require initial value assignment. Whereas Placeholders require just the memory allocation as per the datatype which can be assigned any value later at any point of time using feed_dict.
- Constants : Once a value is assigned to it, then the value cannot be changed during the execution.
- Session : The Graph that we created in the above steps needs to be computed and in TensorFlow the execution for this is done using Session.
- Gradients and Optimizer :
Eq1 :- y = mx + c
Eq2 :- y_given
The main point of this Learning is to find this Eq1. For this we need m and c for getting the equation right, once we have this equation then we are done. Basically our model has to give us Eq1 which gets us y and y_given nearly same. So Gradients is a way of finding these variables and Optimizer is a technique of fine tuning the learnt m and c values.
- Managing Graphs : As TensorFlow is all about creating Computational Graphs and executing them. Hence we have lot of freedom to change from default Graph to another graph which we can see below.
How to use TensorBoard for Visualizing Computation Graphs of TensorFlow?
- Used for Visualizing the tensorflow computational Graph
- Understand the different Scaler Processes
Trigger Tensorboard :- tensorboard –logdir log_dir_name/
After executing the above command just move to the URL that pops up on your terminal and hence you can see the tensorboard. In case the URL is not working then try to use localhost instead of what popped up on your terminal.
Code available on github :-