Use whichever platform you prefer to work on – we will give references and hints for R and for Python.

Do one of the following tasks:

Pick 5000 instances of the same digit from MNIST.

Task

Perform an ICA on this subsample.

Also do a k-Means clustering on the same data with some k you choose yourself.

Look at the k means and compare to the top k independent components. What do you notice?

Task

Perform MDS on this subsample.

What do you notice? Pick out representatives for different regions and compare them. What does the placement of the MDS representatives tell you about the digits?