Setting up deep learning tools/packages in R

The major deep learning packages are developed in C/C++ for efficiency purposes and wrappers are developed in R to efficiently develop, extend, and execute deep learning models.

A lot of open source deep learning libraries are available. The prominent libraries in this area are as follows:

• Theano
• TensorFlow
• Torch
• Caffe

There are other prominent packages available on the market such as H2O, CNTK (Microsoft Cognitive Toolkit), darch, Mocha, and ConvNetJS. There are a lot of wrappers that are developed around these packages to support the easy development of deep learning models, such as Keras and Lasagne in Python and MXNet, both supporting multiple languages.

How to do it...

1. This chapter will cover the MXNet and TensorFlow packages (developed in C++ and CUDA for a highly optimized performance in GPU).
2. Additionally, the h2o package will be used to develop some deep learning models. The h2o package in R is implemented as a REST API, which connects to the H2O server (it runs as Java Virtual Machines (JVM)). We will provide quick setup instructions for these packages in the following sections