If you want to learn tensorflow, this is how you start. In this guide we will show you all the resources you need to learn TensorFlow and start work on your AI projects and deep learning algorithms in no time. In fact, if you have been putting off learning TensorFlow for a while, it is time to start. Tomorrow is not the right time to learn because it never comes: today is. If you are unsure if you need to learn TensorFlow, well, you do need to learn it. So, stop being in that analysis paralysis and start learning now.
What Is TensorFlow?
Before we start to discuss the best way to learn tensorflow, let’s spend a few words on what TensorFlow is. You might know this already, or have a general idea, but I encourage you to read this section anyway because it will act as the basis for the discussion on learning TensorFlow.
We can start with a simple definition of TensorFlow:
TensorFlow is an open-source framework created by Google that implements Deep Learning algorithms.
In other words, it is a Python library that you can use to leverage Deep Learning to solve problems. If you want to learn deep learning, you should learn TensorFlow. Of course, deep learning is a branch of machine learning. But what is that exactly?
Deep learning is the field of Machine Learning that leverage neural networks.
And this, in turns, leads us to the definition of neural network
A neural network is a logical construct where data passes through nodes (neurons) that manipulate it. Neurons are connected with each other, and each connection has a numeric weight.
With deep learning and neural network, you make a computer solve a problem without even realizing specifically how it is doing so. And it can solve complex, unstructured problems that normal algorithms or even traditional machine learning (without neural networks) cannot solve. How cool is that? Yes, you should learn TensorFlow today.
Machine learning and deep learning rely on quite sophisticated mathematics. However, all this is implemented under the hood in TensorFlow, so you don’t need to know the specifics (although it can certainly be useful).
How to Learn TensorFlow
Create a Practice Environment
You cannot just learn TensorFlow by reading documentation or watching videos. The only true way to learn TensorFlow is by practicing what you learn, writing code, running it and seeing if it works or not. This is true in any programming discipline, but especially in Deep Learning where you have a lot of experiments and trial-and-error as part of the process.
TensorFlow is a python library, so you should have a Python environment on your machine. It works on both Windows and Mac, so you may want to check out how to install Python on Windows or how to install Python on Mac guides. It is not that complicated.
However, deep learning performs significantly better on graphic cards (GPUs) rather than normal processors (CPU). You should have a GPU on your machine to be fast in your experiments. But even if you don’t, you don’t need to panic. There is a way to run all your experiments on a GPU online and for free.
If you don’t have a GPU, or if you are struggling to install Python or TensorFlow itself for whatever reason, let me introduce you to Google Colab. Google Colab is a platform where you create “notebooks”, that are documents where you can write text and code, and run that code. Colab comes with a python environment preinstalled, and it has already all the libraries you might need, including TensorFlow, ready to go. Not bad uh?
I personally started to learn TensorFlow on Google Colab before moving onto my machine, and I recommend that to anyone. It is an easy way to get you started and remove the friction you may experience when starting something new.
Learn TensorFlow Online
The best way to learn TensorFlow is online, and it is also quite cheap. You don’t need to spend tons of money to go to on-site courses. There are both paid (but cheap) and totally free options. Here are a few suggestions.
You can start with this long and awesome video by freeCodeCamp.org. The people at Free Code Camp make beautiful content, and this video is a great way to jumpstart your TensorFlow learning and sense if it is something you like. At almost 7 hours, it is complete enough yet relatively short so that you can approach it and complete it in a week. It is completely free.
If after that you realize that Deep Learning is your thing, you should continue to learn TensorFlow on Udemy with this amazing course by Andrei Neagoie and Daniel Bourke. This is quite a complete course, and approaches things slow with a lot of trial and error. With over 60 hours of content, it is your go-to course if you want to learn TensorFlow. You can normally find it on Udemy between 10$ and 20$ (USD).
Now, to be fair, this course is not exhaustive. It won’t uncover 100% of TensorFlow content, but that is because it is too wide of a field, and some things are quite advanced. For example, you will learn how an Algorithm can classify various paragraphs into types (e.g., background vs. conclusion) or how your PC can recognize the picture of a pizza. You won’t learn how to create a chatbot where your computer produces text. But you will have a solid foundation so that you can explore those topics on your own if you want.
Learn TensorFlow by Reading the Docs
Everybody loves to read documentation, right? Well, not quite, but being able to read it is the difference between a good developer and a great developer. Courses cover specific use cases that are most common or popular. The documentation explains how TensorFlow works under the hood, so that you can adjust it to your need.
If you need to go from A to B, you can take an Uber ride and it will get you there. Or, you can rent a car, buy a map, and drive yourself there. If midway you realize you also need to step by C, no problem. Reading the docs is driving yourself, taking the course only is going with the Uber ride.
I suggest to start with the courses because they are a great way to have quick wins and understand the general concepts behind TensorFlow. But when it is time to learn TensorFlow in the true and most effective way, go with the docs. You can find the official documentation here.
Conclusion: Learn TensorFlow Now!
Yeah, without delay, it is time to start to learn TensorFlow now. Don’t postpone this any longer, even 10 minutes are good to start.
As a bonus, if you are not familiar enough with Python, we are giving you access to our complete Python course for free here. Going through online articles, you will learn how to program in Python even if you haven’t seen a single line of code before in your life.
