dl-class: Class 'deeplearning' and sub-classes
In RcppDL: Deep Learning Methods via Rcpp

The class deeplearning is the main class for deep learning methods in this package. It is a virtual class and thus not supposed to be instanciated directly.

The sub-classes implement four methods: da for denoising autoencoder; sda for stacked denoising autoencoder; rbm for restricted Boltzmann machine; dbn for deep belief net.

Methods currently implemented for da

summary: signature(object = "da"): ...
train: signature(object = "da"): ...
CorruptionLevel: signature(object = "da"): ...
LearningRate: signature(object = "da"): ...
reconstruct: signature(object = "da", test = "matrix"): ...
setCorruptionLevel: signature(object = "da", x = "numeric"): ...
setTrainingEpochs: signature(object = "da", x = "numeric"): ...
setLearningRate: signature(object = "da", x = "numeric"): ...
setHiddenRepresentation: signature(object = "da", x = "numeric"): ...

Methods currently implemented for sda

summary: signature(object = "sda"): ...
pretrain: signature(object = "sda"): ...
finetune: signature(object = "sda"): ...
predict: signature(object = "sda", test = "matrix"): ...
CorruptionLevel: signature(object = "sda"): ...
LearningRate: signature(object = "sda"): ...
FinetuneEpochs: signature(object = "sda"): ...
FinetuneLearningRate: signature(object = "sda"): ...
setCorruptionLevel: signature(object = "sda", x = "numeric"): ...
setFinetuneEpochs: signature(object = "sda", x = "numeric"): ...
setFinetuneLearningRate: signature(object = "sda", x = "numeric"): ...
setPretrainLearningRate: signature(object = "sda", x = "numeric"): ...
setPretrainEpochs: signature(object = "sda", x = "numeric"): ...

Methods currently implemented for rbm

summary: signature(object = "rbm"): ...
train: signature(object = "rbm"): ...
LearningRate: signature(object = "rbm"): ...
reconstruct: signature(object = "rbm", test = "matrix"): ...
setStep: signature(object = "rbm", x = "numeric"): ...
setHiddenRepresentation: signature(object = "rbm", x = "numeric"): ...
setLearningRate: signature(object = "rbm", x = "numeric"): ...
setTrainingEpochs: signature(object = "rbm", x = "numeric"): ...

Methods currently implemented for dbn

summary: signature(object = "dbn"): ...
pretrain: signature(object = "dbn"): ...
finetune: signature(object = "dbn"): ...
predict: signature(object = "dbn", test = "matrix"): ...
LearningRate: signature(object = "dbn"): ...
FinetuneEpochs: signature(object = "dbn"): ...
FinetuneLearningRate: signature(object = "dbn"): ...
setFinetuneEpochs: signature(object = "dbn", x = "numeric"): ...
setFinetuneLearningRate: signature(object = "dbn", x = "numeric"): ...
setStep: signature(object = "dbn", x = "numeric"): ...
setPretrainEpochs: signature(object = "dbn", x = "numeric"): ...
setPretrainLearningRate: signature(object = "dbn", x = "numeric"): ...

Qiang Kou

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