probable

Easy and reasonably efficient probabilistic programming and random generation

http://github.com/alpmestan/probable

Latest on Hackage:0.1.3@rev:3

This package is not currently in any snapshots. If you're interested in using it, we recommend adding it to Stackage Nightly. Doing so will make builds more reliable, and allow stackage.org to host generated Haddocks.

BSD-3-Clause licensed by Alp Mestanogullari
Maintained by [email protected]

Easy and reasonably efficient probabilistic programming and random generation

This library gives a common language to speak about probability distributions and random generation, by wrapping both, when necessary, in a RandT monad defined in Math.Probable.Random. This module also provides a lot of useful little combinators for easily describing how random values for your types should be generated.

In Math.Probable.Distribution, you'll find functions for generating random values that follow any distribution supported by mwc-random.

In Math.Probable.Distribution.Finite, you'll find an adaptation of Eric Kidd's work on probability monads (from here).

You may want to check the examples bundled with this package, viewable online at https://github.com/alpmestan/probable/tree/master/examples. One of these examples is simple enough to be worth reproducing here.

module Main where

import Control.Applicative
import Control.Monad
import Math.Probable

import qualified Data.Vector.Unboxed as VU

data Person = Person Int    -- ^ age
                     Double -- ^ weight (kgs)
                     Double -- ^ salary (e.g euros)
    deriving (Eq, Show)

person :: RandT IO Person
person =
    Person <$> uniformIn (1, 100)
           <*> uniformIn (2, 130)
           <*> uniformIn (500, 10000)

randomPersons :: Int -> IO [Person]
randomPersons n = mwc $ listOf n person

randomDoubles :: Int -> IO (VU.Vector Double)
randomDoubles n = mwc $ vectorOf n double

main :: IO ()
main = do
    randomPersons 10 >>= mapM_ print
    randomDoubles 10 >>= VU.mapM_ print

Please report any feature request or problem, either by email or through github's issues/feature requests.