NN.Architectures.RNN

Definitions

RNNCell : Type -> Type -> Type -> Type

  Defines the type of a RNN cell as a parametric map
  @ x the type of the input
  @ s the type of the state
  @ y the type of the output

Totality: total
Visibility: public export

RNN : Type -> Type -> Type -> Nat -> Type

  Defines the type of the unrolled RNN as a parametric map
  @ n the number of unroll steps

Totality: total
Visibility: public export

RNNImpl : (DPair (x, s) (const p) -> (y, s)) -> DPair (Vect n x, s) (const p) -> (Vect n y, s)

  Given a rnn cell, implement the full RNN by iterating that cell
  Helper function for `RNNPara`

Totality: total
Visibility: public export

RNNPara : (cell : RNNCell x s y) -> IsNotDependent cell => RNN x s y n

  Parametric map for the full RNN

Totality: total
Visibility: public export

runRNN : (rnn : RNN x s y n) -> (xs : Vect n x) -> (initialState : s) -> Param rnn (xs, initialState) -> Vect n y

Totality: total
Visibility: public export

exampleRNN : RNNCell Double Double Double

Totality: total
Visibility: public export

exampleInput : Vect 10 Double

Totality: total
Visibility: public export

exampleOutput : Vect 10 Double

Totality: total
Visibility: public export