owl 0.2.9 · OCaml Package

module Neuron : sig ... end

module Optimise : sig ... end

type node = Owl_neural_graph.Make(Owl_dense_matrix.D)(Owl_dense_ndarray.D).node =
  {

mutable name : string;
mutable prev : node array;
mutable next : node array;
mutable neuron : Neuron.neuron;
mutable output : Neuron.t option;
mutable network : network;
mutable train : bool;

}

and network =
  Owl_neural_graph.Make(Owl_dense_matrix.D)(Owl_dense_ndarray.D).network =
  {

mutable nnid : string;
mutable size : int;
mutable root : node option;
mutable topo : node array;

}

val topological_sort : 'a -> 'a option array

val bfs_iter : (node -> 'a) -> node list -> unit

val bfs_map : (node -> 'a) -> node list -> 'a array

val bfs_array : node list -> node array

val make_network : ?nnid:string -> int -> node option -> node array -> network

val make_node : 
  ?name:string ->
  ?train:bool ->
  node array ->
  node array ->
  Neuron.neuron ->
  Neuron.t option ->
  network ->
  node

val get_root : network -> node

val get_node : network -> string -> node

val get_network : node -> network

val collect_output : node array -> Neuron.t array

val connect_pair : node -> node -> unit

val connect_to_parents : node array -> node -> unit

val add_node : 
  ?act_typ:Neuron.Activation.typ ->
  network ->
  node array ->
  node ->
  node

val init : network -> unit

val reset : network -> unit

val mktag : int -> network -> unit

val mkpar : network -> Neuron.t array array

val mkpri : network -> Neuron.t array array

val mkadj : network -> Neuron.t array array

val update : network -> Neuron.t array array -> unit

val run : Neuron.t -> network -> Neuron.t

val forward : network -> Neuron.t -> Neuron.t * Neuron.t array array

val backward : 
  network ->
  Neuron.t ->
  Neuron.t array array * Neuron.t array array

val copy : network -> network

val _remove_training_nodes : network -> unit

val model : network -> Neuron.mat -> Neuron.mat

val model_cnn : network -> Neuron.arr -> Neuron.mat

val input : ?name:string -> int array -> node

val activation : ?name:string -> Neuron.Activation.typ -> node -> node

val linear : 
  ?name:string ->
  ?init_typ:Neuron.Init.typ ->
  ?act_typ:Neuron.Activation.typ ->
  int ->
  node ->
  node

val linear_nobias : 
  ?name:string ->
  ?init_typ:Neuron.Init.typ ->
  ?act_typ:Neuron.Activation.typ ->
  int ->
  node ->
  node

val embedding : 
  ?name:string ->
  ?init_typ:Neuron.Init.typ ->
  ?act_typ:Neuron.Activation.typ ->
  int ->
  int ->
  node ->
  node

val recurrent : 
  ?name:string ->
  ?init_typ:Neuron.Init.typ ->
  act_typ:Neuron.Activation.typ ->
  int ->
  int ->
  node ->
  node

val lstm : ?name:string -> ?init_typ:Neuron.Init.typ -> int -> node -> node

val gru : ?name:string -> ?init_typ:Neuron.Init.typ -> int -> node -> node

val conv1d : 
  ?name:string ->
  ?padding:Owl_types.padding ->
  ?init_typ:Neuron.Init.typ ->
  ?act_typ:Neuron.Activation.typ ->
  int array ->
  int array ->
  node ->
  node

val conv2d : 
  ?name:string ->
  ?padding:Owl_types.padding ->
  ?init_typ:Neuron.Init.typ ->
  ?act_typ:Neuron.Activation.typ ->
  int array ->
  int array ->
  node ->
  node

val conv3d : 
  ?name:string ->
  ?padding:Owl_types.padding ->
  ?init_typ:Neuron.Init.typ ->
  ?act_typ:Neuron.Activation.typ ->
  int array ->
  int array ->
  node ->
  node

val fully_connected : 
  ?name:string ->
  ?init_typ:Neuron.Init.typ ->
  ?act_typ:Neuron.Activation.typ ->
  int ->
  node ->
  node

val max_pool1d : 
  ?name:string ->
  ?padding:Owl_types.padding ->
  ?act_typ:Neuron.Activation.typ ->
  int array ->
  int array ->
  node ->
  node

val max_pool2d : 
  ?name:string ->
  ?padding:Owl_types.padding ->
  ?act_typ:Neuron.Activation.typ ->
  int array ->
  int array ->
  node ->
  node

val avg_pool1d : 
  ?name:string ->
  ?padding:Owl_types.padding ->
  ?act_typ:Neuron.Activation.typ ->
  int array ->
  int array ->
  node ->
  node

val avg_pool2d : 
  ?name:string ->
  ?padding:Owl_types.padding ->
  ?act_typ:Neuron.Activation.typ ->
  int array ->
  int array ->
  node ->
  node

val global_max_pool1d : 
  ?name:string ->
  ?act_typ:Neuron.Activation.typ ->
  node ->
  node

val global_max_pool2d : 
  ?name:string ->
  ?act_typ:Neuron.Activation.typ ->
  node ->
  node

val global_avg_pool1d : 
  ?name:string ->
  ?act_typ:Neuron.Activation.typ ->
  node ->
  node

val global_avg_pool2d : 
  ?name:string ->
  ?act_typ:Neuron.Activation.typ ->
  node ->
  node

val dropout : ?name:string -> float -> node -> node

val gaussian_noise : ?name:string -> float -> node -> node

val gaussian_dropout : ?name:string -> float -> node -> node

val alpha_dropout : ?name:string -> float -> node -> node

val normalisation : 
  ?name:string ->
  ?axis:int ->
  ?training:bool ->
  ?decay:float ->
  ?mu:Neuron.arr ->
  ?var:Neuron.arr ->
  node ->
  node

val reshape : ?name:string -> ?convert:bool -> int array -> node -> node

val flatten : ?name:string -> ?convert:bool -> node -> node

val lambda : 
  ?name:string ->
  ?act_typ:Neuron.Activation.typ ->
  (Neuron.t -> Neuron.t) ->
  node ->
  node

val add : ?name:string -> ?act_typ:Neuron.Activation.typ -> node array -> node

val mul : ?name:string -> ?act_typ:Neuron.Activation.typ -> node array -> node

val dot : ?name:string -> ?act_typ:Neuron.Activation.typ -> node array -> node

val max : ?name:string -> ?act_typ:Neuron.Activation.typ -> node array -> node

val average : 
  ?name:string ->
  ?act_typ:Neuron.Activation.typ ->
  node array ->
  node

val concatenate : 
  ?name:string ->
  ?act_typ:Neuron.Activation.typ ->
  int ->
  node array ->
  node

val to_string : network -> string

val print : network -> unit

val save : network -> string -> unit

val load : string -> network

val save_weights : network -> string -> unit

val load_weights : network -> string -> unit

val train_generic : 
  ?state:Optimise.Checkpoint.state ->
  ?params:Optimise.Params.typ ->
  ?init_model:bool ->
  network ->
  Optimise.t ->
  Optimise.t ->
  Optimise.Checkpoint.state

val train : 
  ?state:Optimise.Checkpoint.state ->
  ?params:Optimise.Params.typ ->
  ?init_model:bool ->
  network ->
  Neuron.mat ->
  Neuron.mat ->
  Optimise.Checkpoint.state

val train_cnn : 
  ?state:Optimise.Checkpoint.state ->
  ?params:Optimise.Params.typ ->
  ?init_model:bool ->
  network ->
  Neuron.arr ->
  Neuron.mat ->
  Optimise.Checkpoint.state

package owl