Getting Started with Tensorflow

It has been almost a year since Tensorflow was released by Google.Although there are a lot of deep learning libraries available(like Theano etc.) but Tensorflow is pretty big!One of the prominent reason is being backed by the big fish,Google! Also tensorflow has pretty great support for distributed systems.Considering the open-source popularity of tensorflow and recent advancements in neural network research,this library is here to stay. In this post we will not only introduce tensorflow but also take a under-the-hood trip to its working.We will start off by going through basics of using tensorflow and analyze "computational graphs" that form the basis of tensorflow's working.Later we will build a linear regression model that would further clarify its working. When we come across the name "Tensorflow",the first thing that invariably comes to mind is the word "tensor".Why "tensor"flow?What is a "tensor"?Well,not dwelling too much on its mathematical representation,consider tensor as a multidimensional array of numbers.Thus all scalars,vectors,matrices fall under the category of tensors. In above program the function tf.constant(value) is used to declare a constant of value value and tf.add(a,b) is used to add two tensors a and b.