diff --git a/.gitignore b/.gitignore
--- a/.gitignore
+++ b/.gitignore
@@ -11,3 +11,4 @@
 *.hi
 *~
 *#
+*.imports
diff --git a/.travis.yml b/.travis.yml
--- a/.travis.yml
+++ b/.travis.yml
@@ -7,11 +7,14 @@
   - travis/cabal-apt-install $mode
 
 install:
+  - cabal install packunused packdeps
   - cabal configure -flib-Werror $mode
-  - cabal build
+  - cabal build --ghc-options=-ddump-minimal-imports
 
 script:
   - $script
+  - packdeps ad.cabal
+  - packunused
   - hlint src --cpp-define HLINT --cpp-include include
 
 notifications:
diff --git a/CHANGELOG.markdown b/CHANGELOG.markdown
--- a/CHANGELOG.markdown
+++ b/CHANGELOG.markdown
@@ -1,3 +1,7 @@
+4.2.1
+-----
+* Added `stochasticGradientDescent`.
+
 4.2
 ---
 * Removed broken `Directed` mode.
diff --git a/ad.cabal b/ad.cabal
--- a/ad.cabal
+++ b/ad.cabal
@@ -1,5 +1,5 @@
 name:         ad
-version:      4.2.0.1
+version:      4.2.1
 license:      BSD3
 license-File: LICENSE
 copyright:    (c) Edward Kmett 2010-2014,
@@ -109,13 +109,13 @@
     data-reify       >= 0.6   && < 0.7,
     erf              >= 2.0   && < 2.1,
     free             >= 4.6.1 && < 5,
-    mtl              >= 2     && < 2.2,
     nats             >= 0.1.2 && < 1,
     reflection       >= 1.4   && < 2,
-    tagged           >= 0.7   && < 1,
-    template-haskell,
-    transformers     >= 0.3   && < 0.4
+    transformers     >= 0.3   && < 0.5
 
+  if impl(ghc < 7.8)
+    build-depends: tagged >= 0.7 && < 1
+
   exposed-modules:
     Numeric.AD
     Numeric.AD.Halley
@@ -167,8 +167,7 @@
     base,
     directory,
     doctest >= 0.9.0.1 && <= 0.10,
-    filepath,
-    mtl
+    filepath
   ghc-options: -Wall -threaded
   if impl(ghc<7.6)
     ghc-options: -Werror
diff --git a/src/Numeric/AD.hs b/src/Numeric/AD.hs
--- a/src/Numeric/AD.hs
+++ b/src/Numeric/AD.hs
@@ -127,6 +127,7 @@
   , gradientAscent
   , conjugateGradientDescent
   , conjugateGradientAscent
+  , stochasticGradientDescent
   ) where
 
 import Data.Functor.Compose
diff --git a/src/Numeric/AD/Newton.hs b/src/Numeric/AD/Newton.hs
--- a/src/Numeric/AD/Newton.hs
+++ b/src/Numeric/AD/Newton.hs
@@ -25,6 +25,7 @@
   , gradientAscent
   , conjugateGradientDescent
   , conjugateGradientAscent
+  , stochasticGradientDescent
   ) where
 
 import Data.Foldable (all, sum)
@@ -37,7 +38,7 @@
 import Numeric.AD.Internal.Reverse (Reverse, Tape)
 import Numeric.AD.Internal.Type (AD(..))
 import Numeric.AD.Mode
-import Numeric.AD.Mode.Reverse as Reverse (gradWith')
+import Numeric.AD.Mode.Reverse as Reverse (gradWith, gradWith')
 import Numeric.AD.Rank1.Kahn as Kahn (Kahn, grad)
 import qualified Numeric.AD.Rank1.Newton as Rank1
 import Prelude hiding (all, mapM, sum)
@@ -121,6 +122,30 @@
         x1 = fmap (\(xi,gxi) -> xi - eta * gxi) xgx
         (fx1, xgx1) = Reverse.gradWith' (,) f x1
 {-# INLINE gradientDescent #-}
+
+-- | The 'stochasticGradientDescent' function approximates
+-- the true gradient of the constFunction by a gradient at
+-- a single example. As the algorithm sweeps through the training 
+-- set, it performs the update for each training example.
+--
+-- It uses reverse mode automatic differentiation to compute the gradient
+-- The learning rate is constant through out, and is set to 0.001
+stochasticGradientDescent :: (Traversable f, Fractional a, Ord a) 
+  => (forall s. Reifies s Tape => f (Scalar a) -> f (Reverse s a) -> Reverse s a) 
+  -> [f (Scalar a)]
+  -> f a 
+  -> [f a]
+stochasticGradientDescent errorSingle d0 x0 = go xgx0 0.001 dLeft
+  where
+    dLeft = tail $ cycle d0
+    xgx0 = Reverse.gradWith (,) (errorSingle (head d0)) x0
+    go xgx !eta d
+      | eta ==0       = []
+      | otherwise     = x1 : go xgx1 eta (tail d)
+      where
+        x1 = fmap (\(xi, gxi) -> xi - eta * gxi) xgx
+        (_, xgx1) = Reverse.gradWith' (,) (errorSingle (head d)) x1
+{-# INLINE stochasticGradientDescent #-}
 
 -- | Perform a gradient descent using reverse mode automatic differentiation to compute the gradient.
 gradientAscent :: (Traversable f, Fractional a, Ord a) => (forall s. Reifies s Tape => f (Reverse s a) -> Reverse s a) -> f a -> [f a]