Math.Meanshift

Basic Meanshift routines

meanShift :: Double -> [Vector] -> Vector -> Vector

Calculate the Mean shift for a point in a dataset. This is efficient only when we cannot make an a priori estimate on which points contribute to the mean shift at given location.

meanShiftWindow :: Int -> Window -> Double -> Vector -> Vector

Mean shift with a windowing function. Performing mean shift is more efficient if we can index and calculate only those points that are in the support of our kernel.

Auxiliary functions for iterating the meanshift steps.

fixedPoint :: Eq a => (a -> a) -> a -> [a]

Find a path to the fixed point of a function.

fixedPointE :: Double -> (Vector -> Vector) -> Vector -> [Vector]

Types

type Window = Support -> [Vector]

type Support = (Vector, Double)

(multidimensional) Kernel Density Estimates

kde :: Double -> [Vector] -> Vector -> Double

Kernel density estimate of given points. Uses a normal kernel.