This is an interactive introduction to a popular dimensionality reduction algorithm: t-distributed stochastic neighbor embedding (t-SNE). Developed by Laurens van der Maaten and Geoffrey Hinton, this algorithm has been successfully applied to many real-world datasets. Here, we'll follow the original paper and describe the key mathematical concepts of the method, when applied to a toy dataset (handwritten digits). We'll use Python and the scikit-learn library.

You can run this in a docker container with all the dependencies pre-installed, interactively in your browser, by pulling this image from docker hub.

After boot2docker start:

docker pull oreilly/an-illustrated-introduction-to-the-t-sne-algorithm
docker run -d -p 8888:8888 -p 80:80 oreilly/an-illustrated-introduction-to-the-t-sne-algorithm
open http://$(boot2docker ip)

If the last step doesn't work, run boot2docker ip and visit that address in a browser. Generally, it's http://192.168.59.103.

Note, this won't work if you're running another container that binds to these ports (80 and 8888).

Originally based on this, which became this.

For this, you need to clone down this repo, and then from this directory.

docker build -t oreilly/an-illustrated-introduction-to-the-t-sne-algorithm  .

docker run -d -p 8888:8888 -p 80:80 -v $PWD/public:/var/www/html oreilly/an-illustrated-introduction-to-the-t-sne-algorithm

Assuming you're running boot2docker, now you can visit http://192.168.59.103 in your browser.

docker run  -p 8888:8888 -p 80:80 -i -t --entrypoint /bin/bash oreilly/an-illustrated-introduction-to-the-t-sne-algorithm