dl-methods: Create 'deeplearning' objects from training set.

In RcppDL: Deep Learning Methods via Rcpp

Description

Rda, Rsda, Rrbm and Rdbn will return an instantiated deeplearning object for denoising autoencoder, stacked denoising autoencoder, restricted Boltzmann machine and deep belief net. train and reconstruct are for training and reconstructing from denoising autoencoder and restricted Boltzmann machine; pretrain, finetune and predict are used for pretraining, finetuning and predicting using stacked denoising autoencoder and deep belief net.

Usage

 Rda(x)
 
 Rsda(x, y, hidden)
 
 Rrbm(x)
 
 Rdbn(x, y, hidden)
 
 train(object)
 
 pretrain(object)
 
 finetune(object)
 
 reconstruct(object, test)
 
 predict(object, test)

Arguments

x	The training dataset.
y	The labels for training dataset.
test	The testing dataset.
hidden	The number of hidden representation in each layer.
object	An instantiated deeplearning object.

Author(s)

Qiang Kou

Examples

 data(test)
 dbn_test <- Rdbn(train_X, train_Y, hidden)
 summary(dbn_test)
 LearningRate(dbn_test)
 pretrain(dbn_test)
 finetune(dbn_test)
 predict(dbn_test, test_X)

Example output

Attaching package: 'RcppDL'

The following object is masked from 'package:stats':

    predict

$PretrainLearningRate
[1] 0.1

$PretrainingEpochs
[1] 1000

$FinetuneLearningRate
[1] 0.1

$FinetuneEpochs
[1] 500

$ContrastiveDivergenceStep
[1] 1

$PretrainLearningRate
[1] 0.1

$FinetuneLearningRate
[1] 0.1

            [,1]        [,2]
[1,] 0.997371818 0.002628182
[2,] 0.003426181 0.996573819

RcppDL documentation built on May 1, 2019, 10:54 p.m.