Module Sig.Normalisation

type neuron_typ = {
  1. mutable axis : int;
  2. mutable beta : Optimise.Algodiff.t;
  3. mutable gamma : Optimise.Algodiff.t;
  4. mutable mu : Optimise.Algodiff.t;
  5. mutable var : Optimise.Algodiff.t;
  6. mutable decay : Optimise.Algodiff.t;
  7. mutable training : bool;
  8. mutable in_shape : int array;
  9. mutable out_shape : int array;
}

Neuron type definition.

val create : ?training:bool -> ?decay:float -> ?mu:Optimise.Algodiff.A.arr -> ?var:Optimise.Algodiff.A.arr -> int -> neuron_typ

Create the neuron. Note that axis 0 is the batch axis.

val connect : int array -> neuron_typ -> unit

Connect this neuron to others in a neural network.

val init : neuron_typ -> unit

Initialise the neuron and its parameters.

val reset : neuron_typ -> unit

Reset the parameters in a neuron.

val mktag : int -> neuron_typ -> unit

Tag the neuron, used by Algodiff module.

val mkpar : neuron_typ -> Optimise.Algodiff.t array

Assemble all the trainable parameters in an array, used by Optimise module.

val mkpri : neuron_typ -> Optimise.Algodiff.t array

Assemble all the primial values in an array, used by Optimise module.

val mkadj : neuron_typ -> Optimise.Algodiff.t array

Assemble all the adjacent values in an array, used by Optimise module.

val update : neuron_typ -> Optimise.Algodiff.t array -> unit

Update trainable parameters of the neuron, used by Optimise module.

val load_weights : neuron_typ -> Optimise.Algodiff.t array -> unit

Load both trainable and non-trainable parameters into the neuron.

val save_weights : neuron_typ -> Optimise.Algodiff.t array

Assemble both trainable and non-trainable parameters of the neuron.

val copy : neuron_typ -> neuron_typ

Make a deep copy of the neuron and its parameters.

Execute the computation in this neuron.

val to_string : neuron_typ -> string

Convert the neuron to its string representation. The string is often a summary of the parameters defined in the neuron.

val to_name : unit -> string

Return the name of the neuron.