diff --git a/ChangeLog.md b/ChangeLog.md
new file mode 100644
--- /dev/null
+++ b/ChangeLog.md
@@ -0,0 +1,3 @@
+# Changelog for Learning
+
+## Unreleased changes
diff --git a/LICENSE b/LICENSE
new file mode 100644
--- /dev/null
+++ b/LICENSE
@@ -0,0 +1,30 @@
+Copyright Bogdan Penkovsky (c) 2018
+
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+    * Redistributions of source code must retain the above copyright
+      notice, this list of conditions and the following disclaimer.
+
+    * Redistributions in binary form must reproduce the above
+      copyright notice, this list of conditions and the following
+      disclaimer in the documentation and/or other materials provided
+      with the distribution.
+
+    * Neither the name of Bogdan Penkovsky nor the names of other
+      contributors may be used to endorse or promote products derived
+      from this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
diff --git a/Learning.cabal b/Learning.cabal
new file mode 100644
--- /dev/null
+++ b/Learning.cabal
@@ -0,0 +1,70 @@
+-- This file has been generated from package.yaml by hpack version 0.20.0.
+--
+-- see: https://github.com/sol/hpack
+--
+-- hash: cc0645cca2baee4a686a5bc4eebc313fda168d1efe6b0301bccb1ec7e7543c25
+
+name:           Learning
+version:        0.0.0
+synopsis:       Most frequently used machine learning tools
+description:    Please see the README on Github at <https://github.com/masterdezign/Learning#readme>
+category:       ML
+homepage:       https://github.com/masterdezign/Learning#readme
+bug-reports:    https://github.com/masterdezign/Learning/issues
+author:         Bogdan Penkovsky
+maintainer:     dev at penkovsky [dot] com
+copyright:      Bogdan Penkovsky
+license:        BSD3
+license-file:   LICENSE
+build-type:     Simple
+cabal-version:  >= 1.10
+
+extra-source-files:
+    ChangeLog.md
+    README.md
+
+source-repository head
+  type: git
+  location: https://github.com/masterdezign/Learning
+
+library
+  hs-source-dirs:
+      src
+  build-depends:
+      base >=4.7 && <5
+    , hmatrix >=0.18.0.0
+    , vector
+  exposed-modules:
+      Learning
+  other-modules:
+      Paths_Learning
+  default-language: Haskell2010
+
+executable Learning-exe
+  main-is: Main.hs
+  hs-source-dirs:
+      app
+  ghc-options: -threaded -rtsopts -with-rtsopts=-N
+  build-depends:
+      Learning
+    , base >=4.7 && <5
+    , hmatrix >=0.18.0.0
+    , vector
+  other-modules:
+      Paths_Learning
+  default-language: Haskell2010
+
+test-suite Learning-test
+  type: exitcode-stdio-1.0
+  main-is: Spec.hs
+  hs-source-dirs:
+      test
+  ghc-options: -threaded -rtsopts -with-rtsopts=-N
+  build-depends:
+      Learning
+    , base >=4.7 && <5
+    , hmatrix >=0.18.0.0
+    , vector
+  other-modules:
+      Paths_Learning
+  default-language: Haskell2010
diff --git a/README.md b/README.md
new file mode 100644
--- /dev/null
+++ b/README.md
@@ -0,0 +1,4 @@
+# Learning
+
+A micro library containing the most common machine learning tools
+written in Haskell.
diff --git a/Setup.hs b/Setup.hs
new file mode 100644
--- /dev/null
+++ b/Setup.hs
@@ -0,0 +1,2 @@
+import Distribution.Simple
+main = defaultMain
diff --git a/app/Main.hs b/app/Main.hs
new file mode 100644
--- /dev/null
+++ b/app/Main.hs
@@ -0,0 +1,4 @@
+module Main where
+
+main :: IO ()
+main = putStrLn "No demo yet"
diff --git a/src/Learning.hs b/src/Learning.hs
new file mode 100644
--- /dev/null
+++ b/src/Learning.hs
@@ -0,0 +1,153 @@
+-- |
+-- = Machine learning utilities
+-- 
+-- A micro library containing the most common machine learning tools.
+-- Check also the mltool package https://hackage.haskell.org/package/mltool.
+
+{-# LANGUAGE UnicodeSyntax #-}
+module Learning (
+  -- * Datasets
+  Dataset (..)
+
+  -- * Principal component analysis
+  , PCA (..)
+  , pca
+
+  -- * Supervised learning
+  , Classifier
+  , learn
+  , learn'
+  , teacher
+  , scores
+  , winnerTakesAll
+
+  -- * Evaluation
+  , errors
+  , errorRate
+  ) where
+
+import           Numeric.LinearAlgebra
+import qualified Data.Vector.Storable as V
+
+-- Supervised dataset
+data Dataset a b = Dataset { _samples :: [a], _labels :: [b] }
+
+-- | Computes "covariance matrix", alternative to (snd. meanCov).
+-- Source: https://hackage.haskell.org/package/mltool-0.1.0.2/docs/src/MachineLearning.PCA.html
+-- covarianceMatrix :: Matrix Double -> Matrix Double
+-- covarianceMatrix x = ((tr x) <> x) / (fromIntegral $ rows x)
+
+-- | Produces a compression matrix u'
+pca' :: Int -> [Vector Double] -> Matrix Double
+pca' maxDim xs = tr u ? [0..maxDim - 1]
+  where
+    xs' = fromBlocks $ map ((: []). tr. reshape 1) xs
+    -- Covariance matrix Sigma
+    sigma = snd $ meanCov xs'
+    -- Eigenvectors matrix U
+    (u, _, _) = svd $ unSym sigma
+
+data PCA = PCA
+  { _u :: Matrix Double  -- Compression matrix U
+  , _compress :: Vector Double -> Matrix Double
+  , _decompress :: Matrix Double -> Vector Double
+  }
+
+-- | Principal component analysis (PCA)
+pca :: Int
+    -> [Vector Double]
+    -> PCA
+pca maxDim xs = let u' = pca' maxDim xs
+                    u = tr u'
+                in PCA
+                   { _u = u
+                   , _compress = (u' <>). reshape 1
+                   , _decompress = flatten. (u <>)
+                   }
+
+type Classifier a = (Matrix Double -> a)
+
+-- | Perform supervised learning to create a linear classifier.
+-- The ridge regression is run with regularization parameter mu=1e-4.
+learn
+  :: V.Storable a =>
+     Vector a
+     -> Matrix Double
+     -> Matrix Double
+     -> Either String (Classifier a)
+learn klasses xs teacher' =
+  case learn' xs teacher' of
+    Just readout -> Right (classify readout klasses)
+    Nothing -> Left "Couldn't learn: check `xs` matrix properties"
+{-# SPECIALIZE learn
+  :: Vector Int
+     -> Matrix Double
+     -> Matrix Double
+     -> Either String (Classifier Int) #-}
+
+-- | Create a linear readout using the ridge regression
+learn'
+  :: Matrix Double
+     -> Matrix Double
+     -> Maybe (Matrix Double)
+learn' a b = case ridgeRegression 1e-4 a b of
+    (Just x) -> Just (tr x)
+    _ -> Nothing
+
+-- | Teacher matrix
+teacher :: Int -> Int -> Int -> Matrix Double
+teacher nLabels correctIndex repeatNo = fromBlocks. map f $ [0..nLabels-1]
+  where ones = konst 1.0 (1, repeatNo)
+        zeros = konst 0.0 (1, repeatNo)
+        f i | i == correctIndex = [ones]
+            | otherwise = [zeros]
+
+-- | Performs the supervised training that results in a linear readout.
+-- See https://en.wikipedia.org/wiki/Tikhonov_regularization
+ridgeRegression :: 
+  Double  -- ^ Regularization constant
+  -> Matrix Double
+  -> Matrix Double 
+  -> Maybe (Matrix Double)
+ridgeRegression μ tA tB = linearSolve oA oB
+  where
+    oA = (tA <> tr tA) + (scalar μ * ident (rows tA))
+    oB = tA <> tr tB
+    _f Nothing = Nothing
+    _f (Just x) = Just (tr x)
+
+-- | Winner-takes-all classification method
+winnerTakesAll
+  :: V.Storable a
+  => Matrix Double  -- ^ Transposed readout matrix
+  -> Vector a  -- ^ Vector of possible classes
+  -> Classifier a  -- ^ `Classifier`
+winnerTakesAll readout klasses response = klasses V.! klass
+  where klass = maxIndex $ scores readout response
+
+-- | Evaluate the network state (nonlinear response) according
+-- to some readout matrix trW.
+scores :: Matrix Double -> Matrix Double -> Vector Double
+scores trW response = evalScores
+  where w = trW <> response
+        -- Sum the elements in each row
+        evalScores = w #> vector (replicate (cols w) 1.0)
+
+classify
+  :: V.Storable a
+     => Matrix Double -> Vector a -> Classifier a
+classify = winnerTakesAll
+{-# SPECIALIZE classify
+  :: Matrix Double -> Vector Int -> Classifier Int
+  #-}
+
+-- | Calculates the error rate in %
+errorRate :: (Eq a, Fractional err) => [a] -> [a] -> err
+errorRate tgtLbls cLbls = 100 * fromIntegral errNo / fromIntegral (length tgtLbls)
+  where errNo = length $ errors $ zip tgtLbls cLbls
+{-# SPECIALIZE errorRate :: [Int] → [Int] → Double #-}
+
+-- | Returns the misclassified cases
+errors :: Eq a => [(a, a)] -> [(a, a)]
+errors = filter (uncurry (/=))
+{-# SPECIALIZE errors :: [(Int, Int)] -> [(Int, Int)] #-}
diff --git a/test/Spec.hs b/test/Spec.hs
new file mode 100644
--- /dev/null
+++ b/test/Spec.hs
@@ -0,0 +1,2 @@
+main :: IO ()
+main = putStrLn "Test suite not yet implemented"