-- Hoogle documentation, generated by Haddock
-- See Hoogle, http://www.haskell.org/hoogle/


-- | Deep Learning in Haskell
--   
--   Implements type-safe deep neural networks
@package deeplearning-hs
@version 0.1.0.2


module DeepLearning.ConvNet

-- | Activation matrix
type Vol sh = Array U sh Double

-- | Delayed activation matrix
type DVol sh = Array D sh Double

-- | Label for supervised learning
type Label = Int

-- | <a>Layer</a> reprsents a layer that can pass activations forward.
--   <a>TopLayer</a> and <a>InnerLayer</a> are derived layers that can be
--   backpropagated through.
class (Shape sh, Shape sh') => Layer a sh sh' | a -> sh, a -> sh'

-- | <a>InnerLayer</a> represents an inner layer of a neural network that
--   can accept backpropagation input from higher layers
class (Layer a sh sh', Shape sh, Shape sh') => InnerLayer a sh sh' | a -> sh, a -> sh'

-- | <a>TopLayer</a> is a top level layer that can initialize a
--   backpropagation pass.
class Layer a DIM1 DIM1 => TopLayer a

-- | <a>SoftMaxLayer</a> computes the softmax activation function.
data SoftMaxLayer
SoftMaxLayer :: SoftMaxLayer

-- | <a>FullyConnectedLayer</a> represents a fully-connected input layer
data FullyConnectedLayer sh
FullyConnectedLayer :: Vol (sh :. Int) -> Vol DIM1 -> FullyConnectedLayer sh
_weights :: FullyConnectedLayer sh -> Vol (sh :. Int)
_bias :: FullyConnectedLayer sh -> Vol DIM1

-- | <a>&gt;-&gt;</a> composes two forward activation functions
(>->) :: (Monad m, Shape sh, Shape sh', Shape sh'') => Forward m sh sh' -> Forward m sh' sh'' -> Forward m sh sh''

-- | The <a>Forward</a> function represents a single forward pass through a
--   layer.
type Forward m sh sh' = Vol sh -> WriterT [Vector Double] m (DVol sh')

-- | <a>withActivations</a> computes the output activation, along with the
--   intermediate activations
withActivations :: Forward m sh sh' -> Vol sh -> m (DVol sh', [Vector Double])

-- | <tt>FlowNetwork</tt> builds a network of the form
--   
--   <pre>
--   Input Layer              Output Softmax
--      +--+
--      |  |   Inner Layers    +--+   +--+
--      |  |                   |  |   |  |
--      |  |   +-+   +-+  +-+  |  |   |  |
--      |  +---+ +---+ +--+ +--+  +---&gt;  |
--      |  |   +-+   +-+  +-+  |  |   |  |
--      |  |                   |  |   |  |
--      |  |                   +--+   +--+
--      +--+
--   </pre>
flowNetwork :: (Monad m, Shape sh) => sh -> Int -> Int -> Int -> Forward m sh DIM1

-- | <a>net1</a> constructs a single-layer fully connected perceptron with
--   softmax output.
net1 :: (Monad m, InnerLayer a sh DIM1, TopLayer a1) => a -> a1 -> Forward m sh DIM1

-- | <a>net1</a> constructs a two-layer fully connected MLP with softmax
--   output.
net2 :: (Monad m, InnerLayer a sh sh', InnerLayer a1 sh' DIM1, TopLayer a2) => a -> a1 -> a2 -> Forward m sh DIM1

-- | <a>newFC</a> constructs a new fully connected layer
newFC :: Shape sh => sh -> Int -> FullyConnectedLayer sh
instance Shape sh => InnerLayer (FullyConnectedLayer sh) sh DIM1
instance Shape sh => Layer (FullyConnectedLayer sh) sh DIM1
instance TopLayer SoftMaxLayer
instance Layer SoftMaxLayer DIM1 DIM1


module DeepLearning.Util

-- | Sample 3x3 matrix used for demonstrations and tests
testShape :: (Z :. Int) :. Int

-- | Random 3x3 matrix
testInput :: Shape sh => sh -> Array U sh Double

-- | Random single-layer network
testNet :: (Monad m, Shape sh) => sh -> Int -> Forward m sh DIM1