The toolset

This part was especially confusing. I was hearing and reading about quite a few names out there.

  • Tensorflow
  • Keras
  • PyTorch
  • Theano
  • Scikit-learn
  • Numpy
  • R
  • Azure ML cloud
  • iPython notebook / Jupyter notebook

Surely I am missing a few names. It was hard for me to narrow down what exactly did I need to get going. Everyone had an opinion and then some.

I use MacOS at home and a Windows 10 computer at work, so I wanted to have something that could work everywhere easily. I had never really worked with Python before so I decided to go in that direction.

Installing Python was easy but then attempting to install Tensorflow wasn’t as simple as I thought. I really liked pip and thought it would be a breeze but was disappointed. Then I ventured into working with the TF docker image but I didn’t really like working with it. I wanted something seamless so let it be for a bit.

I definitely have to mention here that the differences in Python 2 and 3 didn’t help the situation much as well so make sure you read up on the differences. Probably the main one that I have to deal with during my playtime is changing the print statement.

Tumbling through, I finally found Anaconda/Spyder. It is fairly easy to work with, you can also have Jupiter notebook and install additional libraries through the Anaconda interface is usually trivial. Worst-case scenario, you have to open the Anaconda terminal to install a library with the pip command.

Somethings to note about the other stuff out there. Keras is basically a wrapper library built on top of Tensorflow and other lower level frameworks. I think it is a great starting point for beginners when it comes to deep learning.

For simpler stuff like linear regression, we can just use Scikit-learn.

We will also use pandas for data manipulations like loading data from a CSV file.


Get Anaconda. Keras with Tensorflow for deep learning.

Leave a comment