tSNE in 1D is supposed to give interesting results — trying to reproduce http://gis.stackexchange.com/questions/15528/one-dimensional-map-of-the-world/15567#15567
forked from Fil‘s block: t-SNE with Levenshtein distances