STS Cluster Simulation about similaritylearning HOT 4 CLOSED

Already have first results. Might zoom in the interval [3.5, 4.5] to look for the most promising threshold.
The code is a bit ugly because it repeats some functions from other classes, so it might be good to do some refactoring.

from similaritylearning.

@hbredin @saharghannay blocking this until we can discuss about cluster sizes and possible alternatives to get bigger clusters.

from similaritylearning.

Will do violin plots to visualize cluster distribution for each threshold value, we need more information to answer this.

from similaritylearning.

Closing as the priority for clustering-based STS models is now the lowest.
Will reopen or create a new issue when (if) we have the time to do these experiments

from similaritylearning.