rnn : Recurrent Library for Torch
This is an incremental contribution, offering a reusable library for researchers and practitioners using Torch to build recurrent neural networks.
The authors developed the rnn package, a recurrent neural network library for Torch, which provides flexible components for implementing various RNN architectures and was validated by comparing it against existing implementations from two published papers.
The rnn package provides components for implementing a wide range of Recurrent Neural Networks. It is built withing the framework of the Torch distribution for use with the nn package. The components have evolved from 3 iterations, each adding to the flexibility and capability of the package. All component modules inherit either the AbstractRecurrent or AbstractSequencer classes. Strong unit testing, continued backwards compatibility and access to supporting material are the principles followed during its development. The package is compared against existing implementations of two published papers.