MXNet is a super-powerful open source deep learning framework that is built to ease the development of deep learning algorithms. It is used to define, train, and deploy deep neural networks. MXNet is lean, flexible, and ultra-scalable, that is, it allows fast model training and supports a flexible programming model with multiple languages. The problem with existing deep learning frameworks, such as Torch7, Theano, and Caffe, is that users need to learn another system or a different programming flavor.

However, MXNet resolves this issue by supporting multiple languages, such as C++, Python, R, Julia, and Perl. This eliminates the need for users to learn a new language; therefore, they can use the framework and simplify network definitions. MXNet models are able to fit in small amounts of memory and they can be trained on CPUs, GPUs, and on multiple machines with ease. The mxnet package is readily available for the R language and the details of the install can be looked up in Apache Incubator at https://mxnet.incubator.apache.org.