At the Meetup, we'll be using Python 3 and numpy for about 2/3 of the evening. The last third will use Keras. You can use either the Theano or Tensorflow backend.
We'll be working out of a text editor and the IPython console. Starter scripts are found in this repo, and they are written in such a way that you could put blocks in Jupyter Notebooks and evaluate them there instead.
-
Install keras using the command
pip install keras -
In the IPython shell, type
import keras. It will fail (it can't find thetensorflowbackend), but the whole point is to create the file at~/.keras/keras.jsonthat you need to edit to use thetheanobackend if you are using Theano. Read about that here.
This talk is based on resources from iamtrask, Andrey Karpathy, and Jakob Aungiers. Many of the scripts in this repo originated from their blogs, repos, and Github pages and you'll find a more detailed citation in the script. Most scripts have been modified - all errors induced (or fixed!) are my own.