→ Click to place a point. → Place them in clusters (as seen on the thumbnail) for the best effect. → Press done, and your points will be mapped from a two dimensional space to a one dimensional space with as little information lost as possible.
Alternate version: https://scratch.mit.edu/projects/467856605 This version works takes some time to calculate, while the other version is instant. The benefit of this version is that it seems to do a better job. It uses an autoencoder neural network. I programmed it to easily generalise to any dimensions.