User-attended k-means and GMM clustering

[qiyunzhu]

As we discussed in the meeting, user-attended k-means clustering may be a useful feature to implement. Specifically, the user can mouse-click around the visually identified bin centers to set the k value and the initial values of the centroids. Then the standard k-means clustering using the expectation maximization algorithm will start, and it should converge more easily and accurately than random starts.

A more ambitious goal is to implement user-attended Gaussian mixture model clustering. They use can click and drag to define the initial shape of each distribution.

Some code that are already implemented in BinaRena can assist with the implementation of this (these) functions.

@nujinuji . Also @AbhinavChede for information.