Documentation	Build Status	Julia	Testing	Paper

RecurrentLayers.jl

RecurrentLayers.jl extends Flux.jl recurrent layers offering by providing implementations of additional recurrent layers not available in base deep learning libraries.

Features 🚀

The package offers multiple layers for Flux.jl. Currently there are 30+ cells implemented, together with multiple higher level implementations:

Short name	Publication venue	Official implementation
AntisymmetricRNN/GatedAntisymmetricRNN	ICLR 2019	–
ATR	EMNLP 2018	bzhangGo/ATR
BR/BRC	PLOS ONE 2021	nvecoven/BRC
CFN	ICLR 2017	–
coRNN	ICLR 2021	tk-rusch/coRNN
FastRNN/FastGRNN	NeurIPS 2018	Microsoft/EdgeML
IndRNN	CVPR 2018	Sunnydreamrain/IndRNN_Theano_Lasagne
JANET	arXiv 2018	JosvanderWesthuizen/janet
LEM	ICLR 2022	tk-rusch/LEM
LiGRU	IEEE Transactions on Emerging Topics in Computing 2018	mravanelli/theano-kaldi-rnn
LightRU	MDPI Electronics 2023	–
MinimalRNN	NeurIPS 2017	–
MultiplicativeLSTM	Workshop ICLR 2017	benkrause/mLSTM
MGU	International Journal of Automation and Computing 2016	–
MUT1/MUT2/MUT3	ICML 2015	–
NAS	arXiv 2016	tensorflow_addons/rnn
OriginalLSTM	Neural Computation 1997	-
PeepholeLSTM	JMLR 2002	–
RAN	arXiv 2017	kentonl/ran
RHN	ICML 2017	jzilly/RecurrentHighwayNetworks
SCRN	ICLR 2015	facebookarchive/SCRNNs
SGRN	IET 2018	–
STAR	IEEE Transactions on Pattern Analysis and Machine Intelligence 2022	0zgur0/STAckable-Recurrent-network
Typed RNN / GRU / LSTM	ICML 2016	–
UGRNN	ICLR 2017	-
UnICORNN	ICML 2021	tk-rusch/unicornn
WMCLSTM	Neural Networks 2021	–

• Additional wrappers: Stacked RNNs, Multiplicative RNN, and FastSlow.

Installation 💻

You can install RecurrentLayers using either of:

using Pkg
Pkg.add("RecurrentLayers")

julia> ]
pkg> add RecurrentLayers

Getting started 🛠️

The workflow is identical to any recurrent Flux layer: just plug in a new recurrent layer in your workflow and test it out!

Citation

If you use RecurrentLayers.jl in your work, please consider citing

@misc{martinuzzi2025unified,
  doi = {10.48550/ARXIV.2510.21252},
  url = {https://arxiv.org/abs/2510.21252},
  author = {Martinuzzi,  Francesco},
  keywords = {Machine Learning (cs.LG),  Software Engineering (cs.SE),  FOS: Computer and information sciences,  FOS: Computer and information sciences},
  title = {Unified Implementations of Recurrent Neural Networks in Multiple Deep Learning Frameworks},
  publisher = {arXiv},
  year = {2025},
  copyright = {Creative Commons Attribution 4.0 International}
}

License 📜

This project is licensed under the MIT License, except for nas_cell.jl, which is licensed under the Apache License, Version 2.0.

• nas_cell.jl is a reimplementation of the NASCell from TensorFlow and is licensed under the Apache License 2.0. See the file header and LICENSE-APACHE for details.
• All other files are licensed under the MIT License. See LICENSE-MIT for details.

See also

LuxRecurrentLayers.jl: Equivalent library, providing recurrent layers for Lux.jl.

torchrecurrent: Recurrent layers for Pytorch.

ReservoirComputing.jl: Reservoir computing utilities for scientific machine learning. Essentially gradient free trained neural networks.