module.RNN: R6 class representing an RNN.

In modello: Homemade Deep Learning Library

Description

R6 class representing an RNN.

R6 class representing an RNN.

Super class

modello::module -> module.RNN

Methods

Public methods

• module.RNN$new()

• module.RNN$save()

• module.RNN$op()

• module.RNN$pars()

• module.RNN$rus()

• module.RNN$H()

• module.RNN$fc()

• module.RNN$clone()

Inherited methods

• modello::module$j()

• modello::module$name()

• modello::module$npars()

• modello::module$obj()

• modello::module$y()

---

Method new()

Initialisation method

Usage

module.RNN$new(tx, nh, nx, acth, ny, acty, par.h = FALSE)

Arguments

tx

transposition flag. If > 0 op(x) = t(x)

nh

integer vector of length equal to the hidden layes indicating the number of hidden units for each layer

nx

integer, number if inputs

acth

activation function for the recurrent units

ny

integer, number of outputs of the last fully connected layer

acty

activation function for the last fully connected layer

par.h

logical, true => the initial hidden state is treated as a paramter of the optimisation

---

Method save()

Save the module to an RDS file

Usage

module.RNN$save(of, ...)

Arguments

of

output file name or path

...

additional arguments to be passed to the function base::saveRDS

---

Method op()

Runs the calculation stored in the module

Usage

module.RNN$op(X)

Arguments

X

'number', module inputs

---

Method pars()

Returns a list with the model parameters

Usage

module.RNN$pars()

---

Method rus()

Returns a list with the recurrent units

Usage

module.RNN$rus()

---

Method H()

Returns the initial hidden state

Usage

module.RNN$H()

---

Method fc()

Returns the final fully connected layer

Usage

module.RNN$fc()

---

Method clone()

The objects of this class are cloneable with this method.

Usage

module.RNN$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Author(s)

Filippo Monari

modello documentation built on Feb. 2, 2021, 9:06 a.m.