RcppDL (version 0.0.5)

deeplearning-methods: Create deeplearning objects from training set.

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.

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)