An R package to streamline the training, fine-tuning and predicting processes for deep learning. It aims to further simplify the functions in packages such as 'h2o' and 'deepnet'.
library(devtools)
install_github("woobe/deepr")
Extracting hidden features from original predictors (x) and creating new predictors (x_new).
library(deepr)
x <- matrix(rnorm(1000000), nrow = 5000)
dim(x)
[1] 5000 200
1.3 Train a RBM model with 100 hidden features
model_rbm <- train_rbm(x, n_features = 100)
=====================================================================
[deepr]: Training a Restricted Boltzmann Machine
=====================================================================
[deepr]: Removing variables with near zero variance ...
[deepr]: Normalising x to values between 0 and 1 ...
[deepr]: Training a RBM Layer of 10 Hidden Features ...
[deepr]: Returning the RBM Model ...
[deepr]: All Done! Total Duration: 474 sec.
=====================================================================
x_new <- transform_x(model_rbm, x)
=====================================================================
[deepr]: Transforming Predictors Using a Trained RBM
=====================================================================
[deepr]: Pre-processing predictors (x) based on the RBM object ...
[deepr]: Converting original predictors (x) into new ones (x_new) ...
[deepr]: Returning new predictors (x_new) ...
[deepr]: All Done! Total Duration: 1 sec.
=====================================================================
dim(x_new)
[1] 5000 100
## Automatically install the latest version
install_h2o()
## Install a specific version
install_h2o(h2o_ver = "1500")
## Overwrite the current R package
install_h2o(force = TRUE)
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent