R Neural Network package with multiple hidden layers

Question

I have been using Neural Network on matlab, but am not able to locate a package which allows multiple hidden layer for NN. The R Machine Learning Task View suggests the `nnet' package but that allows only one layer networks which are not useful for me.

Are there other packages which can train sophisticated neural networks as the matlab?

(If there are, the CRAN page should probably be updated. Any ideas on how to do that??)

Answer 1

Package neuralnet allows this, you specify a vector with the number of nodes in each hidden layer.

RSNNS (an interface to the Stuttgart Neural Network Simulator) is mentioned on the page you link, and it also allows multiple hidden layers and complex topologies.

As far as updating the Machine Learning & Statistical Learning Task View, you could contact the maintainer, whose name and email address appears at the top of the page.

Answer 2

There are several neural network packages in R:

• Mxnet - LSTM NN & GRU NN & RNN
• RNN - LSTM NN & GRU NN & RNN
• Nnet - Feed-forward neural networks with a single hidden layer
• NeuralNet - backpropagation NN
• TSDyn - Short term forecasting for a univariate time series (NnetTS)
• GMDH - Short term forecasting for a univariate time series using a Group Method of Data Handling neural network
• RSNNS - Stuttgart Neural Network Simulator - high level enough for many architectures. Includes Jordan and Elman neural networks
• H2o.ai - Deep Learning API, back-propagation w/ many
  hyperparameters
• Deepnet - Few deep learning architectures and neural network algorithms, including BP, RBM, DBN
• Tensorflow in R - Deep Learning API, multiple ANNs available