In part three of my k-means clustering jump method write-up, I address initialization methods and visualizing weirdnesses: https://gist.github.com/daveliepmann/67dde9f05df12af134ac344b7a2886d9

@daveliepmann: maybe a strange question given the channel but why did you choose Clojure/Incanter for this project?

@henripal: It's a totally reasonable question—I actually got it at the Clojure Berlin meet-up too! The answer is that this was for fun and I don't enjoy writing R as much as I enjoy writing Clojure. (I do wonder how that will affect readership in the pure data-science population.) Incanter was the logical choice after picking Clojure.