{-# OPTIONS_GHC -fglasgow-exts #-} ----------------------------------------------------------------------------- -- | -- Module : Numeric.GSL.Distribution.Continuous -- Copyright : (c) Alexander Vivian Hugh McPhail 2010 -- License : GPL-style -- -- Maintainer : haskell.vivian.mcphail gmail com -- Stability : provisional -- Portability : uses ffi -- -- GSL continuous random distribution functions -- ----------------------------------------------------------------------------- module Numeric.GSL.Distribution.Continuous ( ZeroParamDist(..), OneParamDist(..) , TwoParamDist(..), ThreeParamDist(..) , MultiParamDist(..) , BivariateDist(..) , DistFunc(..) , random_0p, density_0p , random_1p, density_1p , random_2p, density_2p , random_3p, density_3p , random_mp, density_mp , random_biv, density_biv , spherical_vector ) where ----------------------------------------------------------------------------- import Data.Packed.Vector --import Data.Packed.Matrix hiding(toLists) import Data.Packed.Development --import Numeric.LinearAlgebra.Linear --import Control.Monad(when) import Foreign hiding(shift) --import Foreign.ForeignPtr --import Foreign.Marshal.Alloc(alloca) --import Foreign.C.Types(CInt,CChar) import Foreign.C.Types(CInt) --import Foreign.C.String(newCString,peekCString) --import GHC.ForeignPtr (mallocPlainForeignPtrBytes) --import GHC.Base --import GHC.IOBase --import Prelude hiding(reverse) import Numeric.GSL.Distribution.Common ----------------------------------------------------------------------------- data ZeroParamDist = Landau deriving Enum data OneParamDist = Gaussian -- ^ standard deviation | Exponential -- ^ mean | Laplace -- ^ width | Cauchy -- ^ scale | Rayleigh -- ^ standard deviation | ChiSq -- ^ degrees of freedom | TDist -- ^ degrees of freedom | Logistic -- ^ scale deriving Enum data TwoParamDist = GaussianTail -- ^ limit, standard deviation | ExpPower -- ^ scale, exponent | RayleighTail -- ^ lower limit, standard deviation | Levy -- ^ scale, exponent | Gamma -- ^ par1, par2 | Uniform -- ^ lower, upper | Lognormal -- ^ offset, standard deviation | FDist -- ^ degrees of freedom, degrees of freedom | Beta -- ^ parameter a, parameter b | Pareto -- ^ exponent, scale | Weibull -- ^ scale, exponent | GumbellI -- ^ A, B | GumbellII -- ^ A, B deriving Enum data ThreeParamDist = LevySkew -- ^ scale, exponent, skewness deriving Enum data MultiParamDist = Dirichlet -- ^ size, alpha deriving Enum data BivariateDist = BiGaussian -- ^ standard deviation, standard deviation, correlation coefficient deriving Enum ----------------------------------------------------------------------------- fromei x = fromIntegral (fromEnum x) :: CInt ----------------------------------------------------------------------------- -- | draw a sample from a zero parameter distribution random_0p :: ZeroParamDist -- ^ distribution type -> Int -- ^ random seed -> Double -- ^ result random_0p d s = unsafePerformIO $ distribution_random_zero_param (fromIntegral s) (fromei d) -- | probability of a variate take a value outside the argument density_0p :: ZeroParamDist -- ^ density type -> DistFunc -- ^ distribution function type -> Double -- ^ value -> Double -- ^ result density_0p d f x = unsafePerformIO $ do case d of Landau -> density_only f d x where density_only f' d' x' = if f' /= Density then error "distribution has no CDF" else distribution_dist_zero_param (fromei f') (fromei d') x' foreign import ccall "distribution-aux.h random0" distribution_random_zero_param :: CInt -> CInt -> IO Double foreign import ccall "distribution-aux.h random0_dist" distribution_dist_zero_param :: CInt -> CInt -> Double -> IO Double ----------------------------------------------------------------------------- -- | draw a sample from a one parameter distribution random_1p :: OneParamDist -- ^ distribution type -> Int -- ^ random seed -> Double -- ^ parameter -> Double -- ^ result random_1p d s p = unsafePerformIO $ alloca $ \r -> do check "random1p" $ distribution_random_one_param (fromIntegral s) (fromei d) p r r' <- peek r return r' -- | probability of a variate take a value outside the argument density_1p :: OneParamDist -- ^ density type -> DistFunc -- ^ distribution function type -> Double -- ^ parameter -> Double -- ^ value -> Double -- ^ result density_1p d f p x = unsafePerformIO $ distribution_dist_one_param (fromei f) (fromei d) x p foreign import ccall "distribution-aux.h random1" distribution_random_one_param :: CInt -> CInt -> Double -> Ptr Double -> IO CInt foreign import ccall "distribution-aux.h random1_dist" distribution_dist_one_param :: CInt -> CInt -> Double -> Double -> IO Double ----------------------------------------------------------------------------- -- | draw a sample from a two parameter distribution random_2p :: TwoParamDist -- ^ distribution type -> Int -- ^ random seed -> Double -- ^ parameter 1 -> Double -- ^ parameter 2 -> Double -- ^ result random_2p d s p1 p2 = unsafePerformIO $ alloca $ \r -> do check "random2p" $ distribution_random_two_param (fromIntegral s) (fromei d) p1 p2 r r' <- peek r return r' -- | probability of a variate take a value outside the argument density_2p :: TwoParamDist -- ^ density type -> DistFunc -- ^ distribution function type -> Double -- ^ parameter 1 -> Double -- ^ parameter 2 -> Double -- ^ value -> Double -- ^ result density_2p d f p1 p2 x = unsafePerformIO $ do case d of GaussianTail -> density_only f d p1 p2 x ExpPower -> no_inverse f d p1 p2 x RayleighTail -> density_only f d p1 p2 x Levy -> error "no PDF or CDF for Levy" _ -> distribution_dist_two_param (fromei f) (fromei d) x p1 p2 where density_only f' d' p1' p2' x' = if f' /= Density then error "distribution has no CDF" else distribution_dist_two_param (fromei f') (fromei d') x' p1' p2' no_inverse f' d' p1' p2' x' = if (f' == LowInv || f' == UppInv) then error "distribution has no inverse CDF" else distribution_dist_two_param (fromei f') (fromei d') x' p1' p2' foreign import ccall "distribution-aux.h random2" distribution_random_two_param :: CInt -> CInt -> Double -> Double -> Ptr Double -> IO CInt foreign import ccall "distribution-aux.h random2_dist" distribution_dist_two_param :: CInt -> CInt -> Double -> Double -> Double -> IO Double ----------------------------------------------------------------------------- -- | draw a sample from a three parameter distribution random_3p :: ThreeParamDist -- ^ distribution type -> Int -- ^ random seed -> Double -- ^ parameter 1 -> Double -- ^ parameter 2 -> Double -- ^ parameter 3 -> Double -- ^ result random_3p d s p1 p2 p3 = unsafePerformIO $ alloca $ \r -> do check "random_3p" $ distribution_random_three_param (fromIntegral s) (fromei d) p1 p2 p3 r r' <- peek r return r' -- | probability of a variate take a value outside the argument density_3p :: ThreeParamDist -- ^ density type -> DistFunc -- ^ distribution function type -> Double -- ^ parameter 1 -> Double -- ^ parameter 2 -> Double -- ^ parameter 3 -> Double -- ^ value -> Double -- ^ result density_3p d f p1 p2 p3 x = unsafePerformIO $ do case d of LevySkew -> error "Levy skew has no PDF or CDF" where density_only f' d' p1' p2' p3' x' = if f' /= Density then error "distribution has no CDF" else distribution_dist_three_param (fromei f') (fromei d') x' p1' p2' p3' foreign import ccall "distribution-aux.h random3" distribution_random_three_param :: CInt -> CInt -> Double -> Double -> Double -> Ptr Double -> IO CInt foreign import ccall "distribution-aux.h random3_dist" distribution_dist_three_param :: CInt -> CInt -> Double -> Double -> Double -> Double -> IO Double ----------------------------------------------------------------------------- -- | draw a sample from a three parameter distribution random_mp :: MultiParamDist -- ^ distribution type -> Int -- ^ random seed -> Vector Double -- ^ parameters -> Vector Double -- ^ result random_mp d s p = unsafePerformIO $ do r <- createVector $ dim p app2 (distribution_random_multi_param (fromIntegral s) (fromei d)) vec p vec r "random_mp" return r -- | probability of a variate take a value outside the argument density_mp :: MultiParamDist -- ^ density type -> DistFunc -- ^ distribution function type -> Vector Double -- ^ parameters -> Vector Double -- ^ values -> Double -- ^ result density_mp d f p q = unsafePerformIO $ do case d of Dirichlet -> density_only f d p q where density_only f' d' p' q' = if f' /= Density then error "distribution has no CDF" else alloca $ \r -> do app2 (distribution_dist_multi_param (fromei f') (fromei d') r) vec p vec q "density_mp" r' <- peek r return r' foreign import ccall "distribution-aux.h random_mp" distribution_random_multi_param :: CInt -> CInt -> CInt -> Ptr Double -> CInt -> Ptr Double -> IO CInt foreign import ccall "distribution-aux.h random_mm_dist" distribution_dist_multi_param :: CInt -> CInt -> Ptr Double -> CInt -> Ptr Double -> CInt -> Ptr Double -> IO CInt ----------------------------------------------------------------------------- -- | draw a sample from a bivariate distribution random_biv :: BivariateDist -- ^ distribution type -> Int -- ^ random seed -> Double -- ^ parameter 1 -> Double -- ^ parameter 2 -> Double -- ^ parameter 3 -> (Double,Double) -- ^ result random_biv d s p1 p2 p3 = unsafePerformIO $ alloca $ \r1 -> alloca $ \r2 -> do distribution_random_bivariate (fromIntegral s) (fromei d) p1 p2 p3 r1 r2 r1' <- peek r1 r2' <- peek r2 return (r1',r2') -- | probability of a variate take a value outside the argument density_biv :: BivariateDist -- ^ density type -> DistFunc -- ^ distribution function type -> Double -- ^ parameter 1 -> Double -- ^ parameter 2 -> Double -- ^ parameter 3 -> (Double,Double) -- ^ value -> Double -- ^ result density_biv d f p1 p2 p3 (x,y) = unsafePerformIO $ do case d of BiGaussian -> density_only f d p1 p2 p3 x y where density_only f' d' p1' p2' p3' x' y' = if f' /= Density then error "distribution has no CDF" else distribution_dist_bivariate (fromei f') (fromei d') x' y' p1' p2' p3' foreign import ccall "distribution-aux.h random_biv" distribution_random_bivariate :: CInt -> CInt -> Double -> Double -> Double -> Ptr Double -> Ptr Double -> IO () foreign import ccall "distribution-aux.h random_biv_dist" distribution_dist_bivariate :: CInt -> CInt -> Double -> Double -> Double -> Double -> Double -> IO Double ----------------------------------------------------------------------------- -- | returns a normalised random direction vector from a multivariate gaussian distribution spherical_vector :: Int -- ^ seed -> Int -- ^ vector size -> Vector Double -- result spherical_vector s vs = unsafePerformIO $ do r <- createVector vs app1 (distribution_spherical_vector (fromIntegral s)) vec r "spherical_vector" return r foreign import ccall "distribution-aux.h random_vector" distribution_spherical_vector :: CInt -> CInt -> Ptr Double -> IO CInt ----------------------------------------------------------------------------- ----------------------------------------------------------------------------- -----------------------------------------------------------------------------