hstatistics 0.2.5.4 → 0.3

raw patch · 3 files changed

+44/−34 lines, 3 filesdep ~hmatrixdep ~hmatrix-gsl-statsPVP ok

version bump matches the API change (PVP)

Dependency ranges changed: hmatrix, hmatrix-gsl-stats

API changes (from Hackage documentation)

Files

• CHANGES +4/−0
• hstatistics.cabal +4/−7
• lib/Numeric/Statistics/PCA.hs +36/−27

@@ -109,3 +109,7 @@  0.2.5.4: 		updated for hmatrix 0.18++0.3:+		changed PCA to use SVD as suggested by Pavol Klacansky +		(issue #3)

@@ -1,8 +1,8 @@ Name:               hstatistics-Version:            0.2.5.4+Version:            0.3 License:            BSD3 License-file:       LICENSE-Copyright:          (c) A.V.H. McPhail 2010, 2011, 2012, 2013, 2014, 2016+Copyright:          (c) A.V.H. McPhail 2010--2014, 2016, 2017 Author:             Vivian McPhail Maintainer:         haskell.vivian.mcphail <at> gmail <dot> com Stability:          provisional@@ -30,8 +30,8 @@     Build-Depends:      base >= 4 && < 5,                         array, random,                         vector,-                        hmatrix >= 0.17,-                        hmatrix-gsl-stats >= 0.4+                        hmatrix >= 0.18,+                        hmatrix-gsl-stats >= 0.4.1.6      Extensions:          @@ -46,12 +46,9 @@     other-modules:           C-sources:           -    ghc-prof-options:   -auto-     ghc-options:        -Wall -fno-warn-missing-signatures                               -fno-warn-orphans                               -fno-warn-unused-binds-                        -O2  source-repository head     type:     git

@@ -1,7 +1,7 @@ ----------------------------------------------------------------------------- -- | -- Module      :  Numeric.Statistics.PCA--- Copyright   :  (c) A. V. H. McPhail 2010, 2014+-- Copyright   :  (c) A. V. H. McPhail 2010, 2014, 2017 -- License     :  BSD3 -- -- Maintainer  :  haskell.vivian.mcphail <at> gmail <dot> com@@ -27,38 +27,47 @@  import Numeric.GSL.Statistics -import Numeric.Statistics+--import Numeric.Statistics  -----------------------------------------------------------------------------  -- | find the principal components of multidimensional data greater than --    the threshhold-pca :: I.Array Int (Vector Double)    -- the data+pca :: I.Array Int (Vector Double)     -- the data     -> Double                         -- eigenvalue threshold-    -> Matrix Double+    -> (Vector Double, Matrix Double) -- Eignevalues, Principal components pca d q = let d' = fmap (\x -> x - (scalar $ mean x)) d -- remove the mean from each dimension-              cv = covarianceMatrix d'-              (val',vec') = eigSH $ trustSym cv -- the covariance matrix is real symmetric-              val = toList val'-              vec = toColumns vec'-              v' = zip val vec+              d'' = fromColumns $ I.elems d'+              (_,vec',uni') = svd d''+              vec = toList vec'+              uni = toColumns uni'+              v' = zip vec uni               v = filter (\(x,_) -> x > q) v'  -- keep only eigens > than parameter-          in fromColumns $ snd $ unzip v+              (eigs,vs) = unzip v+          in (fromList eigs,fromColumns vs)   -- | find N greatest principal components of multidimensional data --    according to size of the eigenvalue pcaN :: I.Array Int (Vector Double)    -- the data      -> Int                            -- number of components to return-     -> Matrix Double+     -> (Vector Double, Matrix Double) -- Eignevalues, Principal components pcaN d n = let d' = fmap (\x -> x - (scalar $ mean x)) d -- remove the mean from each dimension-               cv = covarianceMatrix d'-               (val',vec') = eigSH $ trustSym cv  -- the covariance matrix is real symmetric-               val = toList val'-               vec = toColumns vec'-               v' = zip val vec+               d'' = fromColumns $ I.elems d'+               (_,vec',uni') = svd d''+               vec = toList vec'+               uni = toColumns uni'+               v' = zip vec uni                v = take n $ reverse $ sortBy (comparing fst) v'-           in fromColumns $ snd $ unzip v+               (eigs,vs) = unzip v+           in (fromList eigs,fromColumns vs)  +v1 = fromList [1,2,3,4,5,6::Double]+v2 = fromList [2,3,4,5,6,7::Double]+v3 = fromList [3,4,5,6,7,8::Double]++a = fromColumns [v1,v2,v3]+b = I.listArray (1,3::Int) [v1,v2,v3] :: I.Array Int (Vector Double)+                 -- | perform a PCA transform of the original data (remove mean) -- |     Final = M^T Data^T pcaTransform :: I.Array Int (Vector Double)    -- ^ the data@@ -74,11 +83,11 @@           -> I.Array Int (Vector Double)      -- ^ the reduced data pcaReduce d q = let u = fmap (scalar . mean) d                     d' = zipWith (-) (I.elems d) (I.elems u)-                    cv = covarianceMatrix $ I.listArray (I.bounds d) d'-                    (val',vec') = eigSH $ trustSym cv -- the covariance matrix is real symmetric-                    val = toList val'-                    vec = toColumns vec'-                    v' = zip val vec+                    d'' = fromColumns d'+                    (_,vec',uni') = svd d''+                    vec = toList vec'+                    uni = toColumns uni'+                    v' = zip vec uni                     v = filter (\(x,_) -> x > q) v'  -- keep only eigens > than parameter                     m = fromColumns $ snd $ unzip v                  in I.listArray (I.bounds d) $ zipWith (+) (toRows $ m <> (tr' m) <> fromRows d') (I.elems u) @@ -89,11 +98,11 @@            -> I.Array Int (Vector Double)      -- ^ the reduced data, with n principal components pcaReduceN d n = let u = fmap (scalar . mean) d                      d' = zipWith (-) (I.elems d) (I.elems u)-                     cv = covarianceMatrix $ I.listArray (I.bounds d) d'-                     (val',vec') = eigSH $ trustSym cv -- the covariance matrix is real symmetric-                     val = toList val'-                     vec = toColumns vec'-                     v' = zip val vec+                     d'' = fromColumns d'+                     (_,vec',uni') = svd d''+                     vec = toList vec'+                     uni = toColumns uni'+                     v' = zip vec uni                      v = take n $ reverse $ sortBy (comparing fst) v'                      m = fromColumns $ snd $ unzip v                   in I.listArray (I.bounds d) $ zipWith (+) (toRows $ m <> (tr' m) <> fromRows d') (I.elems u)