kylemcdonald/ml-notebook repository overview

⁠ml-notebook

This project is aimed at providing an accessible and reproducible environment for a variety of machine learning toolkits, with a focus on deep learning toolkits. Instead of asking you to follow a set of complex setup instructions, ml-notebook asks you to wait while a tested, pre-built image is installed.

The following tools are available inside the Ubuntu 14.04 image, with Jupyter as an interface:

General:

• matplotlib⁠*
• scipy⁠*
• numpy⁠*
• pyOSC⁠
• Jupyter⁠

Machine Learning:

• gensim⁠
• scikit-learn⁠
• scipy⁠*
• pandas⁠*

Deep Learning:

• Caffe⁠ (with pycaffe)
• Chainer⁠
• TensorFlow⁠
• Theano⁠ (with Keras⁠ and Lasagne⁠)
• Torch⁠

*These are requirements of other libraries, but also interesting in their own right.

⁠Usage

This is only tested this on OSX. Something similar should work on Linux, and possibly Windows with some changes.

1. Install Docker⁠. On Mac, use Docker for Mac.
2. Clone this repository. git clone https://github.com/kylemcdonald/ml-notebook.git && cd ml-notebook
3. (Optional) Run ./update.sh Note: this downloads 2+GB of data and examples. If you just want to look around, browse ml-examples⁠.
4. Run ./run.sh Note: this downloads another 2+GB of data, a pre-built image from Docker.

Type Ctrl-D to exit the ml-notebook Docker. If you accidentally close the Terminal, the Jupyter notebook will keep running in the background. Whenever you want to run the environment again, just call ./run.sh. Calling ./run.sh when ml-notebook is running in the background will restart ml-notebook.

⁠Acknowledgements

Some of the deep learning toolkits are built based on Dockerfiles from Kaixhin⁠.