diff --git a/AI/SVM/Base.hs b/AI/SVM/Base.hs --- a/AI/SVM/Base.hs +++ b/AI/SVM/Base.hs @@ -42,6 +42,7 @@ import Control.Exception import Control.Monad import Data.Binary +import Data.Foldable(Foldable) import Data.IORef import Data.List import Data.Map (Map) @@ -60,7 +61,9 @@ import System.IO.Unsafe import Unsafe.Coerce import qualified Data.ByteString.Lazy as B +import qualified Data.Foldable as F import qualified Data.Map as Map +import qualified Data.Vector.Unboxed as UV import qualified Data.Vector as GV import qualified Data.Vector.Storable as V import qualified Foreign.Concurrent as C @@ -82,6 +85,9 @@ instance SVMVector (GV.Vector Double) where convert = convertDense . GV.convert +instance SVMVector (UV.Vector Double) where + convert = convertDense . UV.convert + instance SVMVector [Double] where convert = convertDense . V.fromList @@ -122,7 +128,7 @@ offsetPtrs addr = take dim [addr `plusPtr` (idx * sizeOf (C'svm_node undefined undefined)) | idx <- scanl (+) 0 lengths] - y = map (realToFrac . fst) v + y = map (double2CDouble . fst) v xs = concatMap (V.toList . snd) v deleteProblem (C'svm_problem l class_array offset_array , node_array) = @@ -186,7 +192,7 @@ realToFrac <$> V.unsafeWith nodes (c'svm_predict modelPtr) -defaultParamers = C'svm_parameter { +defaultParameters = C'svm_parameter { c'svm_parameter'svm_type = c'C_SVC , c'svm_parameter'kernel_type = c'LINEAR , c'svm_parameter'degree = 3 @@ -212,40 +218,40 @@ put cint = put (realToFrac cint :: Double) get = realToFrac <$> (get :: Get Double) -encodeParams C'svm_parameter{..} = do - put c'svm_parameter'svm_type - put c'svm_parameter'kernel_type - put c'svm_parameter'degree - put c'svm_parameter'gamma - put c'svm_parameter'coef0 - put c'svm_parameter'cache_size - put c'svm_parameter'eps - put c'svm_parameter'C - put c'svm_parameter'nr_weight - --put c'svm_parameter'weight_label = nullPtr - --put c'svm_parameter'weight = nullPtr - put c'svm_parameter'nu - put c'svm_parameter'p - put c'svm_parameter'shrinking - put c'svm_parameter'probability +--encodeParams C'svm_parameter{..} = do +-- put c'svm_parameter'svm_type +-- put c'svm_parameter'kernel_type +-- put c'svm_parameter'degree +-- put c'svm_parameter'gamma +-- put c'svm_parameter'coef0 +-- put c'svm_parameter'cache_size +-- put c'svm_parameter'eps +-- put c'svm_parameter'C +-- put c'svm_parameter'nr_weight +-- --put c'svm_parameter'weight_label = nullPtr +-- --put c'svm_parameter'weight = nullPtr +-- put c'svm_parameter'nu +-- put c'svm_parameter'p +-- put c'svm_parameter'shrinking +-- put c'svm_parameter'probability -decodeParams b = do - c'svm_parameter'svm_type <- get - c'svm_parameter'kernel_type <- get - c'svm_parameter'degree <- get - c'svm_parameter'gamma <- get - c'svm_parameter'coef0 <- get - c'svm_parameter'cache_size <- get - c'svm_parameter'eps <- get - c'svm_parameter'C <- get - c'svm_parameter'nr_weight <- get - c'svm_parameter'nu <- get - c'svm_parameter'p <- get - c'svm_parameter'shrinking <- get - c'svm_parameter'probability <- get - let c'svm_parameter'weight_label = nullPtr - c'svm_parameter'weight = nullPtr - return C'svm_parameter{..} +--decodeParams b = do +-- c'svm_parameter'svm_type <- get +-- c'svm_parameter'kernel_type <- get +-- c'svm_parameter'degree <- get +-- c'svm_parameter'gamma <- get +-- c'svm_parameter'coef0 <- get +-- c'svm_parameter'cache_size <- get +-- c'svm_parameter'eps <- get +-- c'svm_parameter'C <- get +-- c'svm_parameter'nr_weight <- get +-- c'svm_parameter'nu <- get +-- c'svm_parameter'p <- get +-- c'svm_parameter'shrinking <- get +-- c'svm_parameter'probability <- get +-- let c'svm_parameter'weight_label = nullPtr +-- c'svm_parameter'weight = nullPtr +-- return C'svm_parameter{..} @@ -304,11 +310,21 @@ ,c'svm_parameter'svm_type=c'NU_SVR} -setParameters svm kernel = parameters - where +withParameters svm kernel ws op = do + ptr_parameters <- malloc + weight_labels <- newArray (map (fromIntegral.fst) ws) + weights <- newArray (map (double2CDouble.snd) ws) + let no_weights = fromIntegral $ length ws + poke ptr_parameters parameters{c'svm_parameter'weight_label=weight_labels,c'svm_parameter'weight=weights,c'svm_parameter'nr_weight=no_weights} + r <- op ptr_parameters + free ptr_parameters + free weights + free weight_labels + return r + where parameters = setTypeParameters svm . setKernelParameters kernel - $ defaultParamers + $ defaultParameters -- Other params that currently cannot be passed: -- epsilon -- termination 0.001 @@ -320,53 +336,52 @@ foreign import ccall "wrapper" wrapPrintF :: (CString -> IO ()) -> IO (FunPtr (CString -> IO ())) +{-#RULES + "listToList" F.toList = id #-} + -- |Create an SVM from the training data -{-#SPECIALIZE trainSVM :: SVMType -> Kernel -> [(Double,SVMNodes)] -> IO (String,SVM) #-} -trainSVM :: (SVMVector a) => SVMType -> Kernel -> [(Double, a)] -> IO (String, SVM) -trainSVM svm kernel (map (second convert) -> dataSet) = do +{-# SPECIALIZE trainSVM :: (SVMVector a) => SVMType -> Kernel -> [(Int,Double)] -> [(Double, a)] -> IO (String, SVM) #-} +{-# SPECIALIZE trainSVM :: (SVMVector a) => SVMType -> Kernel -> GV.Vector (Int,Double) -> GV.Vector (Double, a) -> IO (String, SVM) #-} +trainSVM :: (Foldable f, SVMVector a) => SVMType -> Kernel -> f (Int,Double) -> f (Double, a) -> IO (String, SVM) +trainSVM svm kernel (F.toList -> ws) (map (second convert) . F.toList -> dataSet) = do messages <- newIORef [] let append x = modifyIORef messages (x:) pf <- wrapPrintF (peekCString >=> append) + c'svm_set_print_string_function pf (problem, ptr_nodes) <- createProblem dataSet - ptr_parameters <- malloc - poke ptr_parameters (setParameters svm kernel) - modelPtr <- with problem $ \ptr_problem -> - c'svm_train ptr_problem ptr_parameters - message <- unlines . reverse <$> readIORef messages - (message ,) . SVM <$> C.newForeignPtr modelPtr - (free ptr_parameters - >>deleteProblem (problem, ptr_nodes) - >>modelFinalizer modelPtr) + withParameters svm kernel ws $ \ptr_parameters -> do + modelPtr <- with problem $ \ptr_problem -> + c'svm_train ptr_problem ptr_parameters + message <- unlines . reverse <$> readIORef messages + (message ,) . SVM <$> C.newForeignPtr modelPtr + (deleteProblem (problem, ptr_nodes) + >>modelFinalizer modelPtr) -- |Cross validate SVM. This is faster than training and predicting for each fold -- separately, since there are no extra conversions done between libsvm and haskell. {-#SPECIALIZE crossvalidate :: SVMType -> Kernel -> Int -> [(Double,SVMNodes)] -> IO (String,[Double]) #-} +{-#SPECIALIZE crossvalidate :: SVMType -> Kernel -> Int -> GV.Vector (Double,SVMNodes) -> IO (String,[Double]) #-} crossvalidate - :: (SVMVector b) => SVMType -> Kernel -> Int -> [(Double, b)] -> IO (String, [Double]) -crossvalidate svm kernel folds (map (second convert) -> dataSet) = do + :: (Foldable f, SVMVector b) => SVMType -> Kernel -> Int -> f (Double, b) -> IO (String, [Double]) +crossvalidate svm kernel folds (map (second convert) . F.toList -> dataSet) = do messages <- newIORef [] - let append x = modifyIORef messages (x:) + let append x = return ()-- modifyIORef messages (x:) pf <- wrapPrintF (peekCString >=> append) - -- The above is just a test. Realistically at that point there - -- should be an ioref that captures the output which would then - -- be returned from this function. c'svm_set_print_string_function pf (problem, ptr_nodes) <- createProblem dataSet - ptr_parameters <- malloc - poke ptr_parameters (setParameters svm kernel) - - result_ptr :: Ptr CDouble <- mallocArray (length dataSet) + withParameters svm kernel [] $ \ptr_parameters-> do + result_ptr :: Ptr CDouble <- mallocArray (length dataSet) - with problem $ \ptr_problem -> - c'svm_cross_validation ptr_problem ptr_parameters (fromIntegral folds) result_ptr + with problem $ \ptr_problem -> + c'svm_cross_validation ptr_problem ptr_parameters (fromIntegral folds) result_ptr - res <- peekArray (length dataSet) result_ptr - message <- unlines . reverse <$> readIORef messages + res <- peekArray (length dataSet) result_ptr + message <- unlines . reverse <$> readIORef messages - free result_ptr >> free ptr_parameters >> deleteProblem (problem,ptr_nodes) + free result_ptr >> deleteProblem (problem,ptr_nodes) - return (message,map realToFrac res) + return (message,map realToFrac res) diff --git a/AI/SVM/Simple.hs b/AI/SVM/Simple.hs --- a/AI/SVM/Simple.hs +++ b/AI/SVM/Simple.hs @@ -12,7 +12,7 @@ -- Notes : The module is currently not robust to inputs of wrong dimensionality -- and is affected by security risks inherent in libsvm model loading. -- --- Important TODO-items: +-- Important TODO-items: -- * Handle the issue of crashing the system by passing vectors of dimension to the SVMs -- * Split this library into high and low level parts -- * Saving and loading SVMs @@ -20,7 +20,7 @@ ------------------------------------------------------------------------------- -- This module presents a high level interface for libsvm toolkit. There are three -- main uses cases for it: --- +-- -- 1. You have vectors of reals associated with labels and you wish to assign labels -- to unlabeled vectors. (Classifier machines) -- @@ -39,29 +39,36 @@ ,Kernel(..) ,SVMOneClass(), SVMClassifier(), SVMRegressor() -- * Classifier machines - ,trainClassifier, crossvalidateClassifier, classify + ,trainClassifier, trainWtdClassifier, crossvalidateClassifier, classify + ,chehLin, ChehLinResult(..) -- * One class machines ,trainOneClass, inSet, OneClassResult(..) -- * Regression machines ,trainRegressor, crossvalidateRegressor, predictRegression - -- * Unfortunate utilities + -- * Unfortunate utilities ,Persisting(..) ) where import AI.SVM.Base +import AI.SVM.Common import Control.Applicative -import Control.Arrow (second, (***), (&&&)) +import Control.Arrow (first, second, (***), (&&&)) +import Control.DeepSeq import Control.Monad import Data.Binary +import Data.Foldable (Foldable) +import Data.Function import Data.List import Data.Map (Map) +import Data.Monoid import Data.Tuple +import Foreign.C.Types (CInt(..)) import System.Directory import System.IO.Unsafe +import qualified Control.Monad.Par as P import qualified Data.ByteString.Lazy as B +import qualified Data.Foldable as F import qualified Data.Map as Map -import Foreign.C.Types (CInt) -import AI.SVM.Common -- | Supported SVM classifiers @@ -77,12 +84,16 @@ deriving (Show) data SVMClassifier a = SVMClassifier SVM (Map a Double) (Map Double a) -newtype SVMRegressor = SVMRegressor SVM -newtype SVMOneClass = SVMOneClass SVM +newtype SVMRegressor = SVMRegressor SVM +newtype SVMOneClass = SVMOneClass SVM + +instance NFData a => NFData (SVMClassifier a) where + rnf (SVMClassifier fp m1 m2) = m1 `deepseq` m2 `seq` () + instance (Ord a, Binary a) => Binary (SVMClassifier a) where put (SVMClassifier svm a b) = put svm >> put a >> put b - get = SVMClassifier <$> get <*> get <*> get + get = SVMClassifier <$> get <*> get <*> get instance Binary SVMRegressor where put (SVMRegressor r) = put r @@ -123,38 +134,53 @@ save fp (SVMOneClass a) = saveSVM fp a load fp = SVMOneClass <$> loadSVM fp --- | Train an SVM classifier of given type. +-- | Train an SVM classifier of given type. trainClassifier - :: (SVMVector b, Ord a) => + :: (SVMVector b, Ord a, Foldable f) => ClassifierType -- ^ The type of the classifier -> Kernel -- ^ Kernel - -> [(a, b)] -- ^ Training data + -> f (a, b) -- ^ Training data -> (String, SVMClassifier a) trainClassifier ctype kernel dataset = unsafePerformIO $ do - let (to,from, doubleDataSet) = convertToDouble dataset - (m,svm) <- trainSVM (generalizeClassifier ctype) kernel doubleDataSet + let (to,from, doubleDataSet) = convertToDouble (F.toList dataset) + (m,svm) <- trainSVM (generalizeClassifier ctype) kernel [] doubleDataSet return . (m,) $ SVMClassifier svm to from -convertToDouble dataset = let +-- | Train an SVM classifier of given type. +trainWtdClassifier + :: (Foldable f, SVMVector b, Ord a) => + ClassifierType -- ^ The type of the classifier + -> Kernel -- ^ Kernel + -> f (a, Double) -- ^ Training weights + -> f (a, b) -- ^ Training data + -> (String, SVMClassifier a) +trainWtdClassifier ctype kernel ws dataset = unsafePerformIO $ do + let (to,from, doubleDataSet) = convertToDouble (F.toList dataset) + cw = map (first conv) (F.toList ws) + conv i = round $ to Map.! i + (m,svm) <- trainSVM (generalizeClassifier ctype) kernel cw doubleDataSet + return . (m,) $ SVMClassifier svm to from + +convertToDouble dataset = let l = zip (nub . map fst $ dataset) [1..] to = Map.fromList l from = Map.fromList $ map swap l - in (to,from, map ((to Map.!) *** convert) dataset) + in (to,from, map ((to Map.!) *** convert) dataset) -- | Perform n-foldl cross validation for given set of SVM parameters -crossvalidateClassifier :: (SVMVector b, Ord a) => +crossvalidateClassifier :: (Foldable f, SVMVector b, Ord a) => ClassifierType -- ^ The type of classifier - -> Kernel -- ^ Classifier kernel + -> Kernel -- ^ Classifier kernel -> Int -- ^ Number of folds to use - -> [(a, b)] -- ^ Dataset + -> f (a, b) -- ^ Dataset -> Int -- ^ Seed value. The crossvalidation randomly partitions the data into subsets using this seed value -> (String, [a]) crossvalidateClassifier ctype kernel folds dataset seed = unsafePerformIO $ do - let (to,from, doubleDataSet) = convertToDouble dataset + let (to,from, doubleDataSet) = convertToDouble (F.toList dataset) c_srand (fromIntegral seed) (m,res :: [Double]) <- crossvalidate (generalizeClassifier ctype) kernel folds doubleDataSet return . (m,) $ map (from Map.!) res - where + where labels = map fst @@ -165,8 +191,8 @@ -- | Train an one class classifier trainOneClass :: SVMVector a => Double -> Kernel -> [a] -> (String, SVMOneClass) trainOneClass nu kernel dataset = unsafePerformIO $ do - let doubleDataSet = map (const 1 &&& convert) dataset - (m,svm) <- trainSVM (ONE_CLASS nu) kernel doubleDataSet + let doubleDataSet = map (const 1 &&& convert) dataset + (m,svm) <- trainSVM (ONE_CLASS nu) kernel [] doubleDataSet return . (m,) $ SVMOneClass svm -- | The result type of one class svm. The prediction is that point is either `In`the @@ -175,36 +201,65 @@ -- | Predict wether given point belongs to the region defined by the oneclass svm inSet :: SVMVector a => SVMOneClass -> a -> OneClassResult -inSet (SVMOneClass svm) vector = if predict svm vector <0 +inSet (SVMOneClass svm) vector = if predict svm vector <0 then Out else In -- | Train an SVM regression machine trainRegressor - :: (SVMVector b') => - RegressorType -> Kernel -> [(Double, b')] -> (String, SVMRegressor) + :: (Foldable f, Functor f, SVMVector b') => + RegressorType -> Kernel -> f (Double, b') -> (String, SVMRegressor) trainRegressor rtype kernel dataset = unsafePerformIO $ do - let doubleDataSet = map (second convert) dataset - (m,svm) <- trainSVM (generalizeRegressor rtype) kernel doubleDataSet + let doubleDataSet = fmap (second convert) (F.toList dataset) + (m,svm) <- trainSVM (generalizeRegressor rtype) kernel [] doubleDataSet return . (m,) $ SVMRegressor svm -crossvalidateRegressor :: (SVMVector b) => +crossvalidateRegressor :: (Foldable f, SVMVector b) => RegressorType -- ^ The type of the regressor - -> Kernel -- ^ Kernel + -> Kernel -- ^ Kernel -> Int -- ^ Number of folds to use - -> [(Double, b)] -- ^ Dataset + -> f (Double, b) -- ^ Dataset -> Int -- ^ Seed value. The crossvalidation randomly partitions the data into subsets using this seed value -> (String, [Double]) + crossvalidateRegressor rtype kernel folds dataset seed = unsafePerformIO $ do - let doubleDataSet = map (second convert) dataset + let doubleDataSet = map (second convert) (F.toList dataset) c_srand (fromIntegral seed) (m,res) <- crossvalidate (generalizeRegressor rtype) kernel folds doubleDataSet return (m,res) +data ChehLinResult = Result {cValue, gammaValue, cvAccuracy :: !Double } +instance NFData ChehLinResult + +-- | Train an RBF classifier using crossvalidation and parameter grid search. This is the +-- recommended way of building classifiers for small to medium size datasets. +chehLin :: (Foldable f, SVMVector b, NFData a, Ord a) => + f (a,b) -> (ChehLinResult,SVMClassifier a) +chehLin v = (Result c s a,clf) + where experiments = [ Result c sigma acc + | c <- pows 2 (-5) 15 + , sigma <- pows 2 (-15) 3 + , let res = snd $ crossvalidateClassifier (C c) (RBF sigma) 10 listSet 1231 + , let acc = accuracy trainingClasses res + ] + trainingClasses = map fst . F.toList $ v + eq = uncurry (==) + accuracy as bs = fromIntegral (count eq $ zip as bs) / genericLength as + count :: (Eq a) => (a -> Bool) -> [a] -> Int + count p = length . filter p + listSet = F.toList v + pows base start end = [base ** i | i <- [start..end]] + results = P.runPar . P.parMap id $ experiments + (Result c s a) = maximumBy (compare `on` measure) results + (msg,clf) = trainClassifier (C c) (RBF s) v + +measure (Result _ _ f) = f + + -- | Predict value for given vector via regression predictRegression :: SVMVector a => SVMRegressor -> a -> Double predictRegression (SVMRegressor svm) (convert -> v) = predict svm v - + foreign import ccall "srand" c_srand :: CInt -> IO () diff --git a/Examples/Plot.hs b/Examples/Plot.hs deleted file mode 100644 --- a/Examples/Plot.hs +++ /dev/null @@ -1,39 +0,0 @@ -{-# LANGUAGE ForeignFunctionInterface, BangPatterns, ScopedTypeVariables, TupleSections, - RecordWildCards, NoMonomorphismRestriction #-} -module Main where - -import AI.SVM.Simple -import qualified Data.Vector.Storable as V -import Diagrams.Prelude -import Diagrams.Backend.Cairo.CmdLine -import Diagrams.Backend.Cairo - - - -main = do - let trainingData = [('r', V.fromList [0,0]) - ,('r', V.fromList [1,1]) - ,('b', V.fromList [0,1]) - ,('b', V.fromList [1,0]) - ,('i', V.fromList [0.5,0.5::Double]) - ] - let (m,svm2) = trainClassifier (C 1) (RBF 4) trainingData - let plot = - (circle # fc green # scale 5 5 ) - `atop` - (circle # fc green # scale 5 5 - `atop` circle # scale 100 100 # lineWidth 5) # translate (200,200) - `atop` - (circle # fc green # scale 5 5 # translate (400,400) ) - `atop` - foldl (atop) (circle # scale 1 1) - [circle # scale 5 5 # translate (400*x,400*y) # fc (color svm2 (x,y)) - | x <- [0,0.025..1], y <- [0,0.025..1]] - fst $ renderDia Cairo (CairoOptions ("test.png") (PNG (400,400))) plot - where - color svm (x,y) = case classify svm [x,y] of - 'r' -> red - 'b' -> blue - 'i' -> indigo - -between a x b = a <= x && x <= b diff --git a/Examples/PlotOneClass.hs b/Examples/PlotOneClass.hs deleted file mode 100644 --- a/Examples/PlotOneClass.hs +++ /dev/null @@ -1,37 +0,0 @@ -{-# LANGUAGE ForeignFunctionInterface, BangPatterns, ScopedTypeVariables, TupleSections, - RecordWildCards, NoMonomorphismRestriction #-} -module Main where - -import AI.SVM.Simple -import qualified Data.Vector.Storable as V -import Diagrams.Prelude -import Diagrams.Backend.Cairo.CmdLine -import Diagrams.Backend.Cairo -import System.Random.MWC -import Control.Applicative -import Control.Monad - - -scaledN g = (+0.5) . (/10) <$> normal g - -main = do - pts ::[(Double,Double)] - <- withSystemRandom $ \g -> zip <$> replicateM 30 (scaledN g :: IO Double) - <*> replicateM 30 (scaledN g :: IO Double) - let (msg, svm2) = trainOneClass 0.01 (RBF 1) pts - putStrLn msg - let plot = - foldl (atop) (circle # scale 0.025) - [circle # scale 0.022 # translate (x,y) # fc green - | (x,y) <- pts ] - `atop` - foldl (atop) (circle # scale 0.025) - [circle # scale 0.012 # translate (x,y) # fc (color svm2 (x,y)) - | x <- [0,0.025..1], y <- [0,0.025..1]] - fst $ renderDia Cairo (CairoOptions ("test.png") (PNG (400,400))) (plot # lineWidth 0.002) - where - color svm pt = case inSet svm pt of - In -> red - Out -> black - -between a x b = a <= x && x <= b diff --git a/Examples/SmokeTest.hs b/Examples/SmokeTest.hs deleted file mode 100644 --- a/Examples/SmokeTest.hs +++ /dev/null @@ -1,51 +0,0 @@ -{-# LANGUAGE ForeignFunctionInterface, BangPatterns, ScopedTypeVariables, TupleSections, - RecordWildCards #-} -module Main where - -import AI.SVM.Simple -import qualified Data.Vector.Storable as V - -main = do - svm <- loadSVM "model" - let positiveSample, negativeSample :: V.Vector Double - positiveSample = V.fromList - [0.708333, 1, 1, -0.320755, -0.105023, -1 - , 1, -0.419847, -1, -0.225806, 1, -1] - negativeSample = V.fromList - [0.583333 ,-1 ,0.333333 ,-0.603774 ,1 ,-1 - ,1 ,0.358779 ,-1 ,-0.483871 ,-1 ,1] - - let - pos, neg :: Double - pos = predict svm positiveSample - neg = predict svm negativeSample - print "Testing a loaded model. Expect (1,-1)." - print (pos,neg) - print "Training" - let trainingData = [(-1, V.fromList [0,1]) - ,(-1, V.fromList [1,0]) - ,(1, V.fromList [1,1]) - ,(-1, V.fromList [1,0]) - ,(1, V.fromList [1,1]) - ,(-1, V.fromList [1,0]) - ,(1, V.fromList [1,1]) - ,(-1, V.fromList [1,0]) - ,(1, V.fromList [1,1]) - ,(-1, V.fromList [1,0]) - ,(1, V.fromList [1,1]) - ,(-1, V.fromList [1,0]) - ,(1, V.fromList [1,1]) - ,(-1, V.fromList [1,0]) - ,(1, V.fromList [1,1]) - ,(1, V.fromList [0,0::Double]) - ] - (msg, r) <- crossvalidate (C_SVC 1) (RBF 1) 2 trainingData - print ("cval",msg, r) - - (msg,svm2) <- trainSVM (C_SVC 1) (RBF 1) trainingData - print $ predict svm2 $ [0,1::Double] - print $ predict svm2 $ [1,0::Double] - print $ predict svm2 $ [0.5,0.5::Double] - print $ predict svm2 $ [1,1::Double] - - diff --git a/svm-simple.cabal b/svm-simple.cabal --- a/svm-simple.cabal +++ b/svm-simple.cabal @@ -1,12 +1,15 @@ name: svm-simple -version: 0.2.6 +version: 0.2.7 synopsis: Medium level, simplified, bindings to libsvm description: This is a set of simplified bindings to libsvm suite of support vector machines. This package provides tools for classification, one-class classification and support vector regression. . + Changes in version 0.2.6.1 . + * Bumped dependencies + . Changes in version 0.2.6 . * Fixed a critical bug with training and crossvalidation @@ -44,13 +47,17 @@ build-type: Simple cabal-version: >= 1.8 + +-- Flag benchmark +-- Description: Compile the benchmarking system +-- Default: False -extra-source-files: - Examples/SmokeTest.hs - Examples/Plot.hs - Examples/PlotOneClass.hs library +-- extra-source-files: +-- Examples/SmokeTest.hs +-- Examples/Plot.hs +-- Examples/PlotOneClass.hs other-modules: AI.SVM.Common Exposed-modules: @@ -61,7 +68,25 @@ bytestring >= 0.9.1 && < 0.10, bindings-svm >= 0.2.0 && < 0.3, binary >= 0.5 && < 0.6, - mwc-random >= 0.8 && < 0.9, + mwc-random >= 0.8 && < 0.13, vector >= 0.7.0.1 && < 1.1, - directory >= 1.1.0.0 && < 1.2, - containers >= 0.4.0.0 && < 0.5 + directory >= 1.1.0.0 && < 1.4, + containers >= 0.4.2.0 && < 0.5, + deepseq >= 1.1 && < 1.6, + monad-par >= 0.1.0.3 && < 0.1.1 + +-- executable svm-benchmark +-- if !flag(benchmark) +-- buildable: False +-- main-is: Benchmark.hs +-- build-depends: +-- base >= 4 && < 5, +-- bytestring >= 0.9.1 && < 0.10, +-- bindings-svm >= 0.2.0 && < 0.3, +-- binary >= 0.5 && < 0.6, +-- mwc-random >= 0.8 && < 0.12, +-- vector >= 0.7.0.1 && < 1.1, +-- directory >= 1.1.0.0 && < 1.4, +-- containers >= 0.4.2.0 && < 0.5, +-- deepseq >= 1.1 && < 1.6, +-- criterion >= 0.6 && < 1.0