Pydata Carolinas 2016 Save

:bar_chart: My tutorial for PyData Carolinas 2016

Project README


Here is a link to the talk:

My tutorial for PyData Carolinas 2016. I start out with the basics of Theano and Keras, then dive a bit deeper with word embeddings and recurrent networks, and finish by talking about and visualizing how one type of attention works for RNNs.

Short URL to this page:


This set of activities give an introduction to deep language modeling, based on this blog post. It is ideal for people who are semi-comfortable using Theano or Tensorflow and have some familiarity with Keras, although if you've never tried any of them, the best way to learn is by doing!


The is a script I wrote to easily get it set up on my Mac, but it should work on a Linux system as well. The basic dependencies are Jupyter, Theano, Keras and matplotlib.

Install Pip

Install the Python package manager (if you don't already have it)

sudo easy_install pip # Mac
sudo apt install python-pip # Ubuntu
??? # Windows

Install Virtualenv

Install the virtual environment tool for managing dependencies (if you don't already have it)

sudo pip install virtualenv

Install dependencies

Clone the repository somewhere on your computer and create a virtual environment to install the project dependencies

cd /path/to/desired/directory/
git clone
cd pydata-carolinas-2016
virtualenv venv
source venv/bin/activate
pip install --upgrade Theano keras h5py jupyter matplotlib

Launch the notebooks

You can see the notebooks by launching the Jupyter workspace

ipython notebook


  1. Theano XOR: The "Hello, world!" program. Implement a feed-forward network which learns an XOR function.
  2. Keras XOR: Showing that it is a lot easier and faster to use Keras.
  3. Word Embeddings: Understanding how word embeddings can help.
  4. Recurrent Networks: Building recurrent networks in Theano, and showing how much easier it is in Keras.
  5. Attentional RNNs: Modifying Keras to do special things.


If you want to contribute something, feel free! Just make sure it is grammatically correct and all the code works.

Open Source Agenda is not affiliated with "Pydata Carolinas 2016" Project. README Source: codekansas/pydata-carolinas-2016
Open Issues
Last Commit
6 years ago

Open Source Agenda Badge

Open Source Agenda Rating