diff --git a/sgd.cabal b/sgd.cabal
--- a/sgd.cabal
+++ b/sgd.cabal
@@ -4,10 +4,10 @@
 --
 -- see: https://github.com/sol/hpack
 --
--- hash: 1409d26fec8b15d16a4a06650dc2ce1bf2f51aa72c0e494af222f5c66394ac04
+-- hash: 95a0d8b144a4bd5d9864eca97315546b377194d492e5676678cab11540085447
 
 name:           sgd
-version:        0.8.0.2
+version:        0.8.0.3
 synopsis:       Stochastic gradient descent library
 description:    Import "Numeric.SGD" to use the library.
 category:       Math
@@ -52,7 +52,7 @@
     , data-default >=0.7 && <0.8
     , deepseq >=1.3 && <1.5
     , filepath >=1.3 && <1.5
-    , hmatrix >=0.19 && <0.20
+    , hmatrix >=0.19 && <0.21
     , logfloat >=0.12 && <0.14
     , monad-par >=0.3.4 && <0.4
     , mtl >=2.0 && <2.3
@@ -82,7 +82,7 @@
     , data-default >=0.7 && <0.8
     , deepseq >=1.3 && <1.5
     , filepath >=1.3 && <1.5
-    , hmatrix >=0.19 && <0.20
+    , hmatrix >=0.19 && <0.21
     , logfloat >=0.12 && <0.14
     , monad-par >=0.3.4 && <0.4
     , mtl >=2.0 && <2.3
diff --git a/src/Numeric/SGD.hs b/src/Numeric/SGD.hs
--- a/src/Numeric/SGD.hs
+++ b/src/Numeric/SGD.hs
@@ -366,6 +366,30 @@
     ps -> foldl1' add ps
 
 
+-- -- | Adapt the gradient function to handle (mini-)batches.  The sub-gradients
+-- -- of the individual batch elements are evaluated in parallel based on the
+-- -- given `Strategy`.
+-- batchGradWith
+--   :: (ParamSet p)
+--   => Strategy p
+--   -> (e -> p -> p)
+--   -> ([e] -> p -> p)
+-- batchGradWith strategy grad xs param =
+-- 
+--   addAll grads
+-- 
+--   where
+-- 
+--     groups = partition numCapabilities xs
+--     grads = parMap strategy gradMany groups
+-- 
+--     -- TODO: can we assume here that the group is non-empty?
+--     gradMany = foldl1' add . map (\e -> grad e param)
+-- 
+--     addAll [] = param
+--     addAll ps = foldl1' add ps
+
+
 -- | Adapt the gradient function to handle (mini-)batches.  The function
 -- calculates the individual sub-gradients sequentially.
 batchGradSeq