train_rbm: Extracts hidden features from multiple predictors (x) using...

In woobe/deepr: deepr: Steamlining Deep Learning Procedures in R (Prototype)

Description

Extracts hidden features from multiple predictors (x) using RBM and returns the final RBM model.

Usage

train_rbm(x, n_features = 100, n_batchsize = 100, n_epochs = 100,
  rmv_nearZeroVar = TRUE)

Arguments

x	Multi-column predictors (numeric - matrix).
n_features	Number of hidden features to be extracted. (Default: 100)
n_batchsize	Size of mini batches for RBM training. (Default: 100)
n_epochs	Number of iterative steps required for each RBM training. (Default: 100)
rmv_nearZeroVar	Should the variables of near zero variance to be removed? (Default: TRUE)

Value

rbm The main RBM object with the weights of hidden features and related stats.

nzv The columns that have near zero variance.

pp The caret object that contains the pre-processing stats.

Examples

## Train a single layer of RBM with 100 hidden features
model_rbm <- train_rbm(x, n_features = 100)

## Transform the original predictors (x) into new predictors (x_new)
x_new <- transform_x(model_rbm, x)

woobe/deepr documentation built on May 4, 2019, 9:47 a.m.