This repository contains a refactored version of the assignments of
TensorFlow's Udacity tutorial. Follow the assignment's instructions
to install TensorFlow. Then the only thing you have to do is create
the data
directory, and the rest should work out of the box.
There are still many improvements that could be made to this code, but it makes things slightly clearer than the original IPython notebooks.
In addition to the code refactoring, most solutions are provided. However given the resources at my disposal, I didn't run extensive tests on all assignment. Therefore most of the hyper-parameters still need fine tuning, so don't expect to get state-of-the-art results simply by running the scripts :-).
Please note that most of the problems can probably be solved using a few high-level TensorFlow functions. However the goal here is to dig a bit and find your own solutions.