README

An ML implementation of the Fizz Buzz Kata. You can find the implementation up to feature 1 on the fizz_buzz branch. The master branch contains an implementation of feature 2 ("pop") as well.

Credits

Heavily borrowed from Fizz Buzz Tensorflow, "enhanced" with some integration testing, multiclass prediction and the second fizz buzz feature "pop".

How to run

• Install python 2.7.x: on OSX brew install python, on Linux apt-get python-dev
• Install virtualenv: pip install virtualenv
• Create a virtualenv in the repo root folder with the python 2.7 runtime: virtualenv --python=python2.7 .
• Install dependencies: pip install -r requirements.txt
• Run the tests: ./run-test

Have fun

You can use this as a sandbox for tensorflow to get a feeling for deep learning / neural networks. What happens if...

• you add more data (Increase num_digits)?
• you increase the depth of the network (Increase num_hidden)?
• you allow the network to learn for longer (Increase num_epochs)?
• you change the learning_rate?
• you change the batch_size?
• you change the optimisation method?
• you increase the complexity of the model by adding more or different layers in?

Compile Tensorflow manually to get better speed on CPU

• Here's how to build Tensorflow from source on a Mac
• Use bazel build -c opt --copt=-mavx --copt=-msse4.2 --copt=-msse4.1 --copt=-msse3 --copt=-mavx2 --copt=-mfma -k //tensorflow/tools/pip_package:build_pip_package when building with bazel to get all the advanced CPU instructions