cmaes 0.1.0.1 → 0.1.1

raw patch · 3 files changed

+75/−27 lines, 3 filesdep +randomdep +vectorPVP: major bump suggested

API removals or changes: PVP suggests a major version bump

Dependencies added: random, vector

API changes (from Hackage documentation)

Files

• LICENSE +2/−2
• Numeric/Optimization/Algorithms/CMAES.hs +63/−24
• cmaes.cabal +10/−1

@@ -1,4 +1,4 @@ This package is an aggregate of programs. cma.py (c) Nikolaus Hansen,-2008-2012 is redistributed under GPL 2 or 3. The other programs (c)-Takayuki Muranushi are licensed under MIT license. See the files+2008-2012 is redistributed under GPL 2 or 3. All the other components+(c) Takayuki Muranushi are licensed under MIT license. See the files LICENSE.GPL2, LICENSE.GPL3 and LICENSE.MIT for more details.

@@ -6,16 +6,14 @@  Usage: -(1) create an optimization problem of type @Config@ by one of-    @minimize@, @minimizeIO@ etc.--(2) @run@ it.-+(1) create an optimization problem of type `Config` by one of+    `minimize`, `minimizeIO` etc. +(2) `run` it.  -Let's optimize the following function /f(xs)/. @xs@ is a vector and-@f@ has its minimum at @xs !! i = sqrt(i)@.+Let's optimize the following function /f(xs)/. @xs@ is a list of+Double and @f@ has its minimum at @xs !! i = sqrt(i)@.  >>> import Test.DocTest.Prop >>> let f = sum . zipWith (\i x -> (x*abs x - i)**2) [0..] :: [Double] -> Double@@ -25,10 +23,10 @@  If your optimization is not working well, try: -* Set @scaling@ in the @Config@ to the appropriate search+* Set `scaling` in the `Config` to the appropriate search   range of each parameter. -* Set @tolFun@ in the @Config@ to the appropriate scale of+* Set `tolFun` in the `Config` to the appropriate scale of   the function values.  An example for scaling the function value:@@ -41,11 +39,11 @@  >>> let f3 xs = sum $ zipWith (\i x -> (x*abs x - i)**2) [0,1e100..] xs >>> let xs30 = replicate 10 0 :: [Double]->>> let m3 = (minimize f3 xs30) {scaling = Just (replicate 10 1e50)}+>>> let m3 = (minimize f3 xs30) {scaling = Just (repeat 1e50)} >>> xs31 <- run $ m3 >>> assert $ f3 xs31 / f3 xs30 < 1e-10 -Use @minimizeT@ to optimize functions on traversable structures.+Use `minimizeT` to optimize functions on traversable structures.  >>> import qualified Data.Vector as V >>> let f4 = V.sum . V.imap (\i x -> (x*abs x - fromIntegral i)**2) :: V.Vector Double -> Double@@ -54,7 +52,7 @@   -Or use @minimizeG@ to optimize functions of almost any type. Let's create a triangle ABC+Or use `minimizeG` to optimize functions of almost any type. Let's create a triangle ABC so that AB = 3, AC = 4, BC = 5.  >>> let dist (ax,ay) (bx,by) = ((ax-bx)**2 + (ay-by)**2)**0.5@@ -69,7 +67,19 @@ >>> assert $ abs ((bx-ax)*(cx-ax) + (by-ay)*(cy-ay)) < 1e-10  +When optimizing noisy functions, set `noiseHandling` = @True@ for better results. +>>> import System.Random+>>> let noise = randomRIO (0,1e-2)+>>> let f6Pure = sum . zipWith (\i x -> (x*abs x - i)**2) [0..]+>>> let f6 xs = fmap (f6Pure xs +) noise+>>> xs60 <- run $ (minimizeIO f6 $ replicate 10 0) {noiseHandling = False}+>>> xs61 <- run $ (minimizeIO f6 $ replicate 10 0) {noiseHandling = True}+>>> assert $ f6Pure xs61 < f6Pure xs60++++ -}  @@ -81,6 +91,7 @@ )where  +import           Control.Applicative ((<|>)) import           Control.Monad hiding (forM_, mapM) import qualified Control.Monad.State as State import           Data.Data@@ -112,8 +123,19 @@     -- ^ The global scaling factor.   , scaling       :: Maybe [Double]     -- ^ Typical deviation of each input parameters.+    -- The length of the list is adjusted to be the same as+    -- initXs, e.g. you can lazily use an infinite list here.   , typicalXs     :: Maybe [Double]     -- ^ Typical mean of each input parameters.+    -- The length of this list too, is adjusted to be the same as+    -- initXs.+  , noiseHandling :: Bool+    -- ^ Assume the function to be rugged and/or noisy+  , noiseReEvals  :: Maybe Int+    -- ^ How many re-evaluation to make to estimate the noise.+  , noiseEps      :: Maybe Double+    -- ^ Perturb the parameters by this amount (relative to sigma)+    -- to estimate the noise   , tolFacUpX     :: Maybe Double     -- ^ Terminate when one of the scaling grew too big     -- (initial scaling was too small.)@@ -143,6 +165,9 @@   , sigma0        = 0.25   , scaling       = Nothing   , typicalXs     = Nothing+  , noiseHandling = False+  , noiseReEvals  = Nothing+  , noiseEps      = Just 1e-7   , tolFacUpX     = Just 1e10   , tolUpSigma    = Just 1e20   , tolFun        = Just 1e-11@@ -223,10 +248,24 @@           fail "ohmy god"   loop     where+      probDim :: Int+      probDim = length initXs++      adjustDim :: [a] -> [a] -> [a]+      adjustDim supply orig =+        take probDim $+        catMaybes $+        zipWith (<|>)+          (map Just orig ++ repeat Nothing)+          (map Just supply)+       options :: [(String, String)]       options = concat $ map maybeToList-        [ "scaling_of_variables" `is` scaling-        , "typical_x"            `is` typicalXs+        [ "scaling_of_variables" `is` (fmap$adjustDim [1..] ) scaling+        , "typical_x"            `is` (fmap$adjustDim initXs) typicalXs+        , "noise_handling"       `is` Just noiseHandling+        , "noise_reevals"        `is` noiseReEvals+        , "noise_eps"            `is` noiseEps         , "tolfacupx"            `is` tolFacUpX         , "tolupsigma"           `is` tolUpSigma         , "tolfunhist"           `is` tolFun@@ -295,7 +334,7 @@  -} -getDoubles :: Data d => d -> [Double]+getDoubles :: Data a => a -> [Double] getDoubles d = reverse $ State.execState (everywhereM getter d) []   where     getter :: GenericM (State.State [Double])@@ -304,19 +343,19 @@       let da = fmap (flip asTypeOf (head ys)) $ cast a       case da of         Nothing -> return a-        Just d -> do-          State.put $ d:ys+        Just dd -> do+          State.put $ dd:ys           return a -putDoubles :: Data d => [Double] -> d -> d+putDoubles :: Data a => [Double] -> a -> a putDoubles ys0 d = State.evalState (everywhereM putter d) ys0   where     putter :: GenericM (State.State [Double])-    putter a = do+    putter a0 = do       ys <- State.get-      let ma' = (cast =<<) $ fmap (asTypeOf (head ys)) $ cast a-      case ma' of-        Nothing -> return a-        Just a' -> do+      let ma1 = (cast =<<) $ fmap (asTypeOf (head ys)) $ cast a0+      case ma1 of+        Nothing -> return a0+        Just a1 -> do           State.put $ tail ys-          return a'+          return a1

@@ -1,5 +1,5 @@ name:                cmaes-version:             0.1.0.1+version:             0.1.1 synopsis:            CMA-ES wrapper in Haskell description: @@ -9,7 +9,14 @@   package you need python2 with numpy available on your system. The   package includes @cma.py@ , Nikolaus Hansen's production-level CMA   library: <http://www.lri.fr/~hansen/cmaes_inmatlab.html#python>.+  .+  This package is an aggregate of programs. cma.py (c) Nikolaus+  Hansen, 2008-2012 is redistributed under GPL 2 or 3. All the other+  components (c) Takayuki Muranushi are licensed under MIT+  license. See the files LICENSE.GPL2, LICENSE.GPL3 and LICENSE.MIT+  for more details. + license:             OtherLicense license-file:        LICENSE author:              Takayuki Muranushi@@ -40,7 +47,9 @@                    , doctest-prop >=0.2                    , mtl                      , process+                   , random                    , syb+                   , vector  source-repository head   type:              git