Grenade.Layers.Softmax

Description

Synopsis

Documentation

A Softmax layer

This layer is like a logit layer, but normalises a set of matricies to be probabilities.

One can use this layer as the last layer in a network if they need normalised probabilities.

Constructors

Instances

Show Softmax Source #
Methods showsPrec :: Int -> Softmax -> ShowS # show :: Softmax -> String # showList :: [Softmax] -> ShowS #
Serialize Softmax Source #
Methods put :: Putter Softmax # get :: Get Softmax #
UpdateLayer Softmax Source #
Associated Types type Gradient Softmax :: * Source # Methods runUpdate :: LearningParameters -> Softmax -> Gradient Softmax -> Softmax Source # createRandom :: MonadRandom m => m Softmax Source # runUpdates :: LearningParameters -> Softmax -> [Gradient Softmax] -> Softmax Source #
KnownNat i => Layer Softmax (D1 i) (D1 i) Source #
Associated Types type Tape Softmax (D1 i :: Shape) (D1 i :: Shape) :: * Source # Methods runForwards :: Softmax -> S (D1 i) -> (Tape Softmax (D1 i) (D1 i), S (D1 i)) Source # runBackwards :: Softmax -> Tape Softmax (D1 i) (D1 i) -> S (D1 i) -> (Gradient Softmax, S (D1 i)) Source #
type Gradient Softmax Source #
type Gradient Softmax = ()
type Tape Softmax (D1 i) (D1 i) Source #
type Tape Softmax (D1 i) (D1 i) = S (D1 i)