probable

Easy and reasonably efficient probabilistic programming and random generation/sampling from distributions, based on mwc-random and statistics

Monthly Downloads: 12

Programming language: Haskell

License: BSD 3-clause "New" or "Revised" License

Tags: Math     Statistics    

Latest version: v0.1.3

README

probable

Simple random value generation for haskell, using an efficient random generator and minimizing system calls. But the library also lets you work with distributions over a finite set, adapting code from Eric Kidd's posts, and all the usual distributions covered in the statistics package.

You can see how it looks in examples, or below. You can view the documentation for 0.1 here.

Example

Simple example of random generation for your types, using probable.

module Main where

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

import qualified Data.Vector.Unboxed as VU

data Person = Person 
    { age    :: Int
    , weight :: Double
    , salary :: Int
    } deriving (Eq, Show)

person :: RandT IO Person
person = 
    Person <$> intIn (1, 100)
           <*> doubleIn (2, 130)
           <*> intIn (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

Distributions over finite sets, conditional probabilities and random sampling.

module Main where

import Math.Probable

import qualified Data.Vector as V

data Book = Interesting 
          | Boring
    deriving (Eq, Show)

bookPrior :: Finite d => d Book
bookPrior = weighted [ (Interesting, 0.2) 
                     , (Boring, 0.8) 
                     ]

twoBooks :: Finite d => d (Book, Book)
twoBooks = do
    book1 <- bookPrior
    book2 <- bookPrior
    return (book1, book2)

sampleBooks :: RandT IO (V.Vector Book)
sampleBooks = vectorOf 10 bookPrior

oneInteresting :: Fin (Book, Book)
oneInteresting = bayes $ do
    (b1, b2) <- twoBooks
    condition (b1 == Interesting || b2 == Interesting)
    return (b1, b2)

main :: IO ()
main = do
    print $ exact bookPrior
    mwc sampleBooks >>= print
    print $ exact twoBooks
    print $ exact oneInteresting

Contact

This library is written and maintained by Alp Mestanogullari.

Feel free to contact me for any feedback, comment, suggestion, bug report and what not.