DataSet.Seq: R6 class representign a dataset of sequences to feed RNN

In modello: Homemade Deep Learning Library

Description

R6 class representign a dataset of sequences to feed RNN

R6 class representign a dataset of sequences to feed RNN

Details

It load the data from a R array

Super class

modello::DataSet.Array

Methods

Public methods

• DataSet.Seq$new()

• DataSet.Seq$length()

• DataSet.Seq$feed()

• DataSet.Seq$rewind()

• DataSet.Seq$..pop.index..()

• DataSet.Seq$..feed..()

• DataSet.Seq$..get.data..()

• DataSet.Seq$clone()

Inherited methods

• modello::DataSet.Array$has.next()

• modello::DataSet.Array$reset()

---

Method new()

Initialisation method

Usage

DataSet.Seq$new(X, ls, yi, indexes = NULL)

Arguments

X

matrix/array

ls

integer, sequence length

yi

integer vector, subset of indices to return a targets

indexes

integer vector, a subset of indexes indicating the starting point fir the sequences

---

Method length()

Returns the length of the dataset

Usage

DataSet.Seq$length()

Examples

ds = DataSet.Seq$new(matrix(rnorm(100), 20), 5, 1)
ds$length() # 20

---

Method feed()

Feeds the data into an existing seqience or creating a new one

Usage

DataSet.Seq$feed(x, mode = "head")

Arguments

x

existing sequence, if NULL it creates a new one

mode

chectacter indicating the way to build the sequences

Examples

modello.init(100, 100, 100, 100)
ds = DataSet.Seq$new(matrix(rnorm(100), 20), 5, 1)
SEQ = ds$feed(NULL)
print(names(SEQ))
print(SEQ$x[[1]]$v)
print(SEQ$y$v)
ds$feed(SEQ)
print(SEQ$x[[1]]$v)
print(SEQ$y$v)
modello.close()

---

Method rewind()

Sets the data loader back of n previous states. Only available for "head" and "tail" modes.

Usage

DataSet.Seq$rewind(n, mode = "head")

Arguments

n

number of step to rewind

mode

rewind mode. Possibly the same as the feed mode.

Examples

ds = DataSet.Seq$new(matrix(rnorm(100), 20), 5, 1)
ds$..get.data..("head")
ds$..get.data..("head")
ds$rewind(2)
ds$..get.data..("head")
ds$..get.data..("head")

---

Method ..pop.index..()

Auxiliary method not to be called directly. It pops an index from the dataset according to the given method

Usage

DataSet.Seq$..pop.index..(mode = "head")

Arguments

mode

popping mode

---

Method ..feed..()

Auxiliary method. Not to be called directly.

Usage

DataSet.Seq$..feed..(x, mode)

Arguments

x

existing sequence, if NULL it creates a new one

mode

chectacter indicating the way to build the sequences

---

Method ..get.data..()

Auxiliary method. Not to be called directly.

Usage

DataSet.Seq$..get.data..(mode)

Arguments

mode

chectacter indicating the way to build the sequences

---

Method clone()

The objects of this class are cloneable with this method.

Usage

DataSet.Seq$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Author(s)

Filippo Monari

Examples

## ------------------------------------------------
## Method `DataSet.Seq$length`
## ------------------------------------------------

ds = DataSet.Seq$new(matrix(rnorm(100), 20), 5, 1)
ds$length() # 20

## ------------------------------------------------
## Method `DataSet.Seq$feed`
## ------------------------------------------------

modello.init(100, 100, 100, 100)
ds = DataSet.Seq$new(matrix(rnorm(100), 20), 5, 1)
SEQ = ds$feed(NULL)
print(names(SEQ))
print(SEQ$x[[1]]$v)
print(SEQ$y$v)
ds$feed(SEQ)
print(SEQ$x[[1]]$v)
print(SEQ$y$v)
modello.close()

## ------------------------------------------------
## Method `DataSet.Seq$rewind`
## ------------------------------------------------

ds = DataSet.Seq$new(matrix(rnorm(100), 20), 5, 1)
ds$..get.data..("head")
ds$..get.data..("head")
ds$rewind(2)
ds$..get.data..("head")
ds$..get.data..("head")

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