K-mean for [(d,d)] - slightly improved for readability

kmean.hs

{-# LANGUAGE FlexibleInstances, MultiParamTypeClasses #-}

module Lib
    ( 
      kMeans
    ) where

import GHC.Exts (groupWith)
import Data.List (transpose, sort)
import Data.List.Split (chunksOf)
import Debug.Trace (trace)

type Pt = (Double, Double)

dist (a, b) (c, d) = sqrt $ (a - c) ^ 2 + (b - d) ^ 2

(.:) = (.) (.) (.)

dbg x = trace (show x) x

kMeans xs k n = within same $ take n $ iterate (dbg . centroids . flip cpartition xs) $ dbg $ centroids $ kpartition k xs
    where
        same = all (<0.1) .: zipWith dist
        centroids = sort . fmap center

nearest cs v = snd $ minimum [(dist c v, c) | c<-cs]

cpartition cs xs = groupWith (nearest cs) xs

kpartition n xs = transpose $ chunksOf n xs

center points = (xsum / l, ysum / l)
    where xsum = sum $ fmap fst points
          ysum = sum $ fmap snd points
          l  = fromIntegral $ length points

within _    [x] = x
within same (x:y:xs) | same x y = x
within same (x:xs) = within same xs