This project seeks to streamline the process of installing and experimenting with the current crop of machine learning frameworks/toolkits, namely
The project runs on a Vagrant box and uses Anaconda (actually miniconda) to manage the various Python libraries. The notebooks are served from the VM by Jupyter (see here for some handy tips).
It is very much a work-in-progress. I am currently making my way through Kevin Murphy's Machine Learning: A Probabilistic Perspective. My plan is to create a corresponding notebook for each thematic area covered therein e.g.
- Bayesian models
- Regression
- Kernels
- Markov models
- State space models
- Clustering
- Deep learning
$ git clone [email protected]:philipobrien/machine-learning-basecamp.git
$ cd machine-learning-basecamp
$ vagrant up
Once the machine is ready (it can take quite a while)
$ vagrant ssh
$ cd /vagrant
$ make basecamp
$ source activate basecamp
$ cd /vagrant/notebooks
$ jupyter notebook --ip=0.0.0.0 --no-browser
In your browser navigate to localhost:8888 and have fun!
Currently there are only three demos
- an iTorch demo which just verifies iTorch is working by loading the default image of Grace Hopper
- a Spark demo which executes a basic logistic regression example
- a TensorFlow demo which runs an MNIST example